Nomad changes

docs/snippets/api_server/cli.md

@@ -0,0 +1,11 @@
+```bash title="Bash"
+# Default: http://127.0.0.1:8000
+kreuzberg serve
+
+# Custom host and port
+kreuzberg serve -H 0.0.0.0 -p 3000
+
+# With configuration file
+kreuzberg serve --config kreuzberg.toml
+
+```

docs/snippets/api_server/csharp.md

@@ -0,0 +1,24 @@
+```csharp title="C#"
+using System;
+using System.Diagnostics;
+
+class ApiServer
+{
+    static void Main()
+    {
+        var processInfo = new ProcessStartInfo
+        {
+            FileName = "kreuzberg",
+            Arguments = "serve -H 0.0.0.0 -p 8000",
+            UseShellExecute = false,
+            RedirectStandardOutput = true,
+            RedirectStandardError = true
+        };
+
+        using (var process = Process.Start(processInfo))
+        {
+            process?.WaitForExit();
+        }
+    }
+}
+```

docs/snippets/api_server/docker.md

@@ -0,0 +1,8 @@
+```bash title="Bash"
+# Run server on port 8000
+docker run -d \n  -p 8000:8000 \n  ghcr.io/kreuzberg-dev/kreuzberg:latest \n  serve -H 0.0.0.0 -p 8000
+
+# With environment variables
+docker run -d \n  -e KREUZBERG_CORS_ORIGINS="https://myapp.com" \n  -e KREUZBERG_MAX_MULTIPART_FIELD_BYTES=209715200 \n  -p 8000:8000 \n  ghcr.io/kreuzberg-dev/kreuzberg:latest \n  serve -H 0.0.0.0 -p 8000
+
+```

docs/snippets/api_server/go.md

@@ -0,0 +1,18 @@
+```go title="Go"
+package main
+
+import (
+    "log"
+    "os/exec"
+)
+
+func main() {
+    cmd := exec.Command("kreuzberg", "serve", "-H", "0.0.0.0", "-p", "8000")
+    cmd.Stdout = log.Writer()
+    cmd.Stderr = log.Writer()
+    if err := cmd.Run(); err != nil {
+        log.Fatalf("failed to start server: %v", err)
+    }
+}
+
+```

docs/snippets/api_server/java.md

@@ -0,0 +1,19 @@
+```java title="Java"
+import java.io.IOException;
+
+public class ApiServer {
+    public static void main(String[] args) {
+        try {
+            ProcessBuilder pb = new ProcessBuilder(
+                "kreuzberg", "serve", "-H", "0.0.0.0", "-p", "8000"
+            );
+            pb.inheritIO();
+            Process process = pb.start();
+            process.waitFor();
+        } catch (IOException | InterruptedException e) {
+            System.err.println("Failed to start server: " + e.getMessage());
+        }
+    }
+}
+
+```

docs/snippets/api_server/python.md

@@ -0,0 +1,6 @@
+```python title="Python"
+# Start server
+import subprocess
+subprocess.Popen(["python", "-m", "kreuzberg", "serve", "-H", "0.0.0.0", "-p", "8000"])
+
+```

docs/snippets/api_server/rust.md

@@ -0,0 +1,11 @@
+```rust title="Rust"
+use kreuzberg::{ExtractionConfig, api::serve_with_config};
+
+#[tokio::main]
+async fn main() -> kreuzberg::Result<()> {
+    let config = ExtractionConfig::discover()?;
+    serve_with_config("0.0.0.0", 8000, config).await?;
+    Ok(())
+}
+
+```

docs/snippets/c/advanced/chunk_page_mapping.md

@@ -0,0 +1,50 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 500,"
+        "\"overlap\": 50"
+        "},"
+        "\"pages\": {"
+        "\"extract_pages\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks (JSON, includes per-chunk first_page/last_page metadata):\n%s\n",
+           chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    char *pages_json = kreuzberg_extraction_result_pages(result);
+    printf("pages (JSON): %s\n", pages_json ? pages_json : "[]");
+    kreuzberg_free_string(pages_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/chunking_config.md

@@ -0,0 +1,43 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"markdown\","
+        "\"max_characters\": 500,"
+        "\"overlap\": 50,"
+        "\"prepend_heading_context\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks (JSON): %s\n", chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/chunking_rag.md

@@ -0,0 +1,47 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 500,"
+        "\"overlap\": 50,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("research_paper.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks (JSON, each item includes content, embedding, and metadata.chunk_index/total_chunks/byte_start/byte_end):\n%s\n",
+           chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/embedding_with_chunking.md

@@ -0,0 +1,48 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 1024,"
+        "\"overlap\": 100,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true,"
+        "\"batch_size\": 32,"
+        "\"show_download_progress\": false"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks with embeddings (JSON):\n%s\n", chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/keyword_extraction_config.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"keywords\": {"
+        "\"algorithm\": \"yake\","
+        "\"max_keywords\": 10,"
+        "\"min_score\": 0.3,"
+        "\"ngram_range\": [1, 3],"
+        "\"language\": \"en\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *keywords_json = kreuzberg_extraction_result_extracted_keywords(result);
+    printf("keywords (JSON): %s\n", keywords_json ? keywords_json : "[]");
+    kreuzberg_free_string(keywords_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/keyword_extraction_example.md

@@ -0,0 +1,46 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"keywords\": {"
+        "\"algorithm\": \"yake\","
+        "\"max_keywords\": 10,"
+        "\"min_score\": 0.3"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("research_paper.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *keywords_json = kreuzberg_extraction_result_extracted_keywords(result);
+    if (keywords_json) {
+        printf("Keywords: %s\n", keywords_json);
+        kreuzberg_free_string(keywords_json);
+    } else {
+        printf("Keywords: (none)\n");
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/language_detection_config.md

@@ -0,0 +1,42 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"language_detection\": {"
+        "\"enabled\": true,"
+        "\"min_confidence\": 0.8,"
+        "\"detect_multiple\": false"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *langs_json = kreuzberg_extraction_result_detected_languages(result);
+    printf("detected languages (JSON): %s\n", langs_json ? langs_json : "[]");
+    kreuzberg_free_string(langs_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/language_detection_multilingual.md

@@ -0,0 +1,42 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"language_detection\": {"
+        "\"enabled\": true,"
+        "\"min_confidence\": 0.8,"
+        "\"detect_multiple\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("multilingual_document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *langs_json = kreuzberg_extraction_result_detected_languages(result);
+    printf("Detected languages: %s\n", langs_json ? langs_json : "[]");
+    kreuzberg_free_string(langs_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/quality_processing_config.md

@@ -0,0 +1,37 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"enable_quality_processing\": true"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    double score = kreuzberg_extraction_result_quality_score(result);
+    printf("quality score: %.2f\n", score);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/quality_processing_example.md

@@ -0,0 +1,45 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"enable_quality_processing\": true"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("scanned_document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    double score = kreuzberg_extraction_result_quality_score(result);
+    if (score < 0.5) {
+        printf("Warning: Low quality extraction (%.2f)\n", score);
+    } else {
+        printf("Quality score: %.2f\n", score);
+    }
+
+    char *warnings_json = kreuzberg_extraction_result_processing_warnings(result);
+    printf("processing warnings (JSON): %s\n", warnings_json ? warnings_json : "[]");
+    kreuzberg_free_string(warnings_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/token_reduction_config.md

@@ -0,0 +1,41 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"token_reduction\": {"
+        "\"mode\": \"moderate\","
+        "\"preserve_important_words\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("reduced content:\n%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/token_reduction_example.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"token_reduction\": {"
+        "\"mode\": \"moderate\","
+        "\"preserve_important_words\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("verbose_document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    if (content) {
+        printf("reduced content (%zu bytes):\n%s\n", strlen(content), content);
+        kreuzberg_free_string(content);
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/advanced/vector_database_integration.md

@@ -0,0 +1,55 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *document_path = "document.pdf";
+    const char *document_id = "doc-001";
+
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 512,"
+        "\"overlap\": 50,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true,"
+        "\"batch_size\": 32"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync(document_path, NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    /* The chunks JSON array carries content + embedding + metadata for each
+       chunk. Pass this directly to your vector database client (pgvector,
+       Qdrant, Pinecone, etc.) along with the document_id as a metadata field. */
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("document_id: %s\n", document_id);
+    printf("chunks (JSON, ready to upsert into a vector DB):\n%s\n",
+           chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/batch_extract_bytes_sync.md

@@ -0,0 +1,33 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    /* Items is a JSON array of BatchBytesItem objects.
+     * Each entry has "content" (array of byte integers), "mime_type", and an optional "config". */
+    const char *items_json =
+        "["
+        "  {\"content\": [72,101,108,108,111,33], \"mime_type\": \"text/plain\"},"
+        "  {\"content\": [87,111,114,108,100,33], \"mime_type\": \"text/plain\"}"
+        "]";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    /* Returns a JSON array of ExtractionResult objects, or NULL on failure. */
+    char *results_json =
+        kreuzberg_batch_extract_bytes_sync(items_json, config);
+    if (!results_json) {
+        fprintf(stderr, "batch extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    printf("%s\n", results_json);
+    kreuzberg_free_string(results_json);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/batch_extract_files_sync.md

@@ -0,0 +1,34 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    /* Items is a JSON array of BatchFileItem objects.
+     * Each entry has a "path" field and an optional "config" override. */
+    const char *items_json =
+        "["
+        "  {\"path\": \"doc1.pdf\"},"
+        "  {\"path\": \"doc2.docx\"},"
+        "  {\"path\": \"scan.png\", \"config\": {\"force_ocr\": true}}"
+        "]";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    /* Returns a JSON array of ExtractionResult objects, or NULL on failure. */
+    char *results_json =
+        kreuzberg_batch_extract_files_sync(items_json, config);
+    if (!results_json) {
+        fprintf(stderr, "batch extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    printf("%s\n", results_json);
+    kreuzberg_free_string(results_json);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/client_chunk_text.md

@@ -0,0 +1,73 @@
+<!-- snippet:syntax-only -->
+
+```c title="C"
+#include <curl/curl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+struct response_buffer {
+    char *data;
+    size_t size;
+};
+
+static size_t write_callback(void *contents, size_t size, size_t nmemb, void *userp) {
+    size_t total = size * nmemb;
+    struct response_buffer *buf = (struct response_buffer *)userp;
+    char *resized = realloc(buf->data, buf->size + total + 1);
+    if (!resized) {
+        return 0;
+    }
+    buf->data = resized;
+    memcpy(buf->data + buf->size, contents, total);
+    buf->size += total;
+    buf->data[buf->size] = '\0';
+    return total;
+}
+
+int main(void) {
+    curl_global_init(CURL_GLOBAL_DEFAULT);
+
+    CURL *curl = curl_easy_init();
+    if (!curl) {
+        fprintf(stderr, "curl_easy_init failed\n");
+        curl_global_cleanup();
+        return 1;
+    }
+
+    const char *body =
+        "{"
+        "\"text\": \"Lorem ipsum dolor sit amet, consectetur adipiscing elit.\","
+        "\"chunker_type\": \"character\","
+        "\"config\": {\"max_characters\": 256, \"overlap\": 32, \"trim\": true}"
+        "}";
+
+    struct curl_slist *headers = NULL;
+    headers = curl_slist_append(headers, "Content-Type: application/json");
+    headers = curl_slist_append(headers, "Accept: application/json");
+
+    struct response_buffer response = {NULL, 0};
+
+    curl_easy_setopt(curl, CURLOPT_URL, "http://localhost:8000/chunk");
+    curl_easy_setopt(curl, CURLOPT_HTTPHEADER, headers);
+    curl_easy_setopt(curl, CURLOPT_POSTFIELDS, body);
+    curl_easy_setopt(curl, CURLOPT_POSTFIELDSIZE, (long)strlen(body));
+    curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, write_callback);
+    curl_easy_setopt(curl, CURLOPT_WRITEDATA, &response);
+
+    CURLcode rc = curl_easy_perform(curl);
+    if (rc != CURLE_OK) {
+        fprintf(stderr, "request failed: %s\n", curl_easy_strerror(rc));
+    } else {
+        long status = 0;
+        curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &status);
+        printf("HTTP %ld\n%s\n", status, response.data ? response.data : "(empty)");
+    }
+
+    free(response.data);
+    curl_slist_free_all(headers);
+    curl_easy_cleanup(curl);
+    curl_global_cleanup();
+    return rc == CURLE_OK ? 0 : 1;
+}
+```

docs/snippets/c/api/client_extract_single_file.md

@@ -0,0 +1,65 @@
+<!-- snippet:syntax-only -->
+
+```c title="C"
+#include <curl/curl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+struct response_buffer {
+    char *data;
+    size_t size;
+};
+
+static size_t write_callback(void *contents, size_t size, size_t nmemb, void *userp) {
+    size_t total = size * nmemb;
+    struct response_buffer *buf = (struct response_buffer *)userp;
+    char *resized = realloc(buf->data, buf->size + total + 1);
+    if (!resized) {
+        return 0;
+    }
+    buf->data = resized;
+    memcpy(buf->data + buf->size, contents, total);
+    buf->size += total;
+    buf->data[buf->size] = '\0';
+    return total;
+}
+
+int main(void) {
+    curl_global_init(CURL_GLOBAL_DEFAULT);
+
+    CURL *curl = curl_easy_init();
+    if (!curl) {
+        fprintf(stderr, "curl_easy_init failed\n");
+        curl_global_cleanup();
+        return 1;
+    }
+
+    struct response_buffer response = {NULL, 0};
+
+    curl_mime *form = curl_mime_init(curl);
+    curl_mimepart *part = curl_mime_addpart(form);
+    curl_mime_name(part, "file");
+    curl_mime_filedata(part, "document.pdf");
+
+    curl_easy_setopt(curl, CURLOPT_URL, "http://localhost:8000/extract");
+    curl_easy_setopt(curl, CURLOPT_MIMEPOST, form);
+    curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, write_callback);
+    curl_easy_setopt(curl, CURLOPT_WRITEDATA, &response);
+
+    CURLcode rc = curl_easy_perform(curl);
+    if (rc != CURLE_OK) {
+        fprintf(stderr, "request failed: %s\n", curl_easy_strerror(rc));
+    } else {
+        long status = 0;
+        curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &status);
+        printf("HTTP %ld\n%s\n", status, response.data ? response.data : "(empty)");
+    }
+
+    free(response.data);
+    curl_mime_free(form);
+    curl_easy_cleanup(curl);
+    curl_global_cleanup();
+    return rc == CURLE_OK ? 0 : 1;
+}
+```

docs/snippets/c/api/combining_all_features.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    /* Combine chunking, OCR, image extraction, and Markdown output in one config. */
+    const char *config_json =
+        "{"
+        "\"output_format\": \"markdown\","
+        "\"force_ocr\": true,"
+        "\"ocr\": {\"backend\": \"tesseract\", \"languages\": [\"eng\", \"deu\"]},"
+        "\"chunking\": {\"chunker_type\": \"character\", \"max_characters\": 1024, \"overlap\": 128, \"trim\": true},"
+        "\"images\": {\"extract_images\": true, \"target_dpi\": 300, \"inject_placeholders\": true}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("document.pdf", NULL, config);
+    if (!result) {
+        int32_t code = kreuzberg_last_error_code();
+        const char *message = kreuzberg_last_error_context();
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                code, message ? message : "(no message)");
+        kreuzberg_extraction_config_free(config);
+        return code != 0 ? code : 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/error_handling.md

@@ -0,0 +1,29 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    /* Pass an unsupported MIME type to trigger an error. */
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_bytes_sync(NULL, 0, "application/x-unknown", config);
+    if (!result) {
+        int32_t code = kreuzberg_last_error_code();
+        const char *message = kreuzberg_last_error_context();
+        /* message is valid until the next FFI call on this thread — copy if needed. */
+        fprintf(stderr, "error %d: %s\n", code, message ? message : "(no message)");
+        kreuzberg_extraction_config_free(config);
+        return code != 0 ? code : 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/error_handling_extract.md

@@ -0,0 +1,47 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    /* Mixed-validity batch: a real PDF, a missing file, and an unsupported type. */
+    const char *items_json =
+        "["
+        "  {\"path\": \"document.pdf\"},"
+        "  {\"path\": \"does-not-exist.pdf\"},"
+        "  {\"path\": \"archive.unknownext\"}"
+        "]";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+    if (!config) {
+        fprintf(stderr, "config init failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    /* Returns a JSON array of ExtractionResult objects (one per input, in order),
+     * or NULL on a system-level failure. Per-item errors are encoded inside
+     * each result object's metadata (e.g. an "errors" array). */
+    char *results_json = kreuzberg_batch_extract_files(items_json, config);
+    if (!results_json) {
+        int32_t code = kreuzberg_last_error_code();
+        const char *message = kreuzberg_last_error_context();
+        /* message is valid until the next FFI call on this thread — copy if needed. */
+        fprintf(stderr, "batch extraction aborted (code %d): %s\n",
+                code, message ? message : "(no message)");
+        kreuzberg_extraction_config_free(config);
+        return code != 0 ? code : 1;
+    }
+
+    /* Walk the returned JSON. A real consumer would feed this to a JSON parser
+     * and inspect each result's metadata.errors / content fields. */
+    size_t len = strlen(results_json);
+    printf("results (%zu bytes):\n%s\n", len, results_json);
+
+    kreuzberg_free_string(results_json);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/extract_bytes_async.md

@@ -0,0 +1,35 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* kreuzberg_extract_bytes schedules work on the global Tokio runtime and
+ * returns once extraction is complete.  For true non-blocking use, call it
+ * from a dedicated OS thread and synchronize via a semaphore or callback. */
+int main(void) {
+    const char *text = "Hello, kreuzberg!";
+    const uint8_t *bytes = (const uint8_t *)text;
+    size_t len = strlen(text);
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_bytes(bytes, len, "text/plain", config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/extract_bytes_sync.md

@@ -0,0 +1,32 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    const char *text = "Hello, kreuzberg!";
+    const uint8_t *bytes = (const uint8_t *)text;
+    size_t len = strlen(text);
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_bytes_sync(bytes, len, "text/plain", config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/extract_file_async.md

@@ -0,0 +1,30 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+/* kreuzberg_extract_file schedules work on the global Tokio runtime and
+ * returns once extraction is complete.  For true non-blocking use, call it
+ * from a dedicated OS thread and synchronize via a semaphore or callback. */
+int main(void) {
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/api/extract_file_sync.md

@@ -0,0 +1,27 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/advanced_config.md

@@ -0,0 +1,25 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    struct ConfigBuilder *builder = kreuzberg_config_builder_new();
+    kreuzberg_config_builder_set_use_cache(builder, 1);
+    kreuzberg_config_builder_set_include_document_structure(builder, 1);
+    kreuzberg_config_builder_set_ocr(builder,
+        "{\"tesseract\":{\"language\":\"eng\"}}");
+
+    ExtractionConfig *config = kreuzberg_config_builder_build(builder);
+
+    struct CExtractionResult *result =
+        kreuzberg_extract_file_sync_with_config("scan.pdf",
+            kreuzberg_config_to_json(config));
+    if (result && result->success) {
+        printf("%s\n", result->content);
+    }
+
+    kreuzberg_free_result(result);
+    kreuzberg_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/chunking_config.md

@@ -0,0 +1,43 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 1000,"
+        "\"overlap\": 200,"
+        "\"trim\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/config_basic.md

@@ -0,0 +1,38 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"use_cache\": true,"
+        "\"enable_quality_processing\": true"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/config_discover.md

@@ -0,0 +1,62 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* The C FFI does not expose config-file auto-discovery directly. Load the
+ * file contents in your application and pass the JSON to
+ * kreuzberg_extraction_config_from_json. For TOML/YAML, convert in your
+ * application before calling the FFI. */
+static char *read_text_file(const char *path) {
+    FILE *fp = fopen(path, "rb");
+    if (!fp) {
+        return NULL;
+    }
+    fseek(fp, 0, SEEK_END);
+    long size = ftell(fp);
+    fseek(fp, 0, SEEK_SET);
+    char *buf = (char *)malloc((size_t)size + 1);
+    if (!buf) {
+        fclose(fp);
+        return NULL;
+    }
+    fread(buf, 1, (size_t)size, fp);
+    buf[size] = '\0';
+    fclose(fp);
+    return buf;
+}
+
+int main(void) {
+    char *json = read_text_file("kreuzberg.json");
+    KREUZBERGExtractionConfig *config = json
+        ? kreuzberg_extraction_config_from_json(json)
+        : kreuzberg_extraction_config_default();
+    free(json);
+
+    if (!config) {
+        fprintf(stderr, "config load failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/config_ocr.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"tesseract\","
+        "\"language\": \"eng\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("scanned.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/config_programmatic.md

@@ -0,0 +1,48 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"use_cache\": true,"
+        "\"enable_quality_processing\": true,"
+        "\"ocr\": {"
+        "\"backend\": \"tesseract\","
+        "\"language\": \"eng+deu\","
+        "\"tesseract_config\": {\"psm\": 6}"
+        "},"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 1000,"
+        "\"overlap\": 200"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/document_structure_config.md

@@ -0,0 +1,37 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"include_document_structure\": true"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/element_based_output.md

@@ -0,0 +1,37 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"result_format\": \"element_based\""
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/embedding_config.md

@@ -0,0 +1,47 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 1000,"
+        "\"overlap\": 200,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"batch_size\": 16,"
+        "\"normalize\": true,"
+        "\"show_download_progress\": true"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/html_output.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"output_format\": \"html\","
+        "\"html_output\": {"
+        "\"theme\": \"github\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/keyword_extraction_config.md

@@ -0,0 +1,43 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"keywords\": {"
+        "\"algorithm\": \"yake\","
+        "\"max_keywords\": 10,"
+        "\"min_score\": 0.1,"
+        "\"ngram_range\": [1, 3],"
+        "\"language\": \"en\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/language_detection_config.md

@@ -0,0 +1,41 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"language_detection\": {"
+        "\"enabled\": true,"
+        "\"min_confidence\": 0.8,"
+        "\"detect_multiple\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/ocr_dpi_config.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"images\": {"
+        "\"extract_images\": true,"
+        "\"target_dpi\": 300,"
+        "\"max_image_dimension\": 4096,"
+        "\"auto_adjust_dpi\": true,"
+        "\"min_dpi\": 150,"
+        "\"max_dpi\": 600"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/pdf_config.md

@@ -0,0 +1,41 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"pdf_options\": {"
+        "\"extract_images\": true,"
+        "\"passwords\": [\"password123\"],"
+        "\"extract_metadata\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("encrypted.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/pdf_hierarchy_config.md

@@ -0,0 +1,45 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"pdf_options\": {"
+        "\"hierarchy\": {"
+        "\"enabled\": true,"
+        "\"detection_threshold\": 0.75,"
+        "\"ocr_coverage_threshold\": 0.8,"
+        "\"min_level\": 1,"
+        "\"max_level\": 5"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/postprocessor_config.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"postprocessor\": {"
+        "\"enabled\": true,"
+        "\"enabled_processors\": [\"whitespace_normalizer\", \"unicode_normalizer\"]"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/quality_processing_config.md

@@ -0,0 +1,38 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"enable_quality_processing\": true,"
+        "\"use_cache\": true"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/tesseract_config.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"tesseract\","
+        "\"language\": \"eng+deu\","
+        "\"tesseract_config\": {"
+        "\"psm\": 6,"
+        "\"oem\": 3"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("scanned.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/config/token_reduction_config.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"token_reduction\": {"
+        "\"mode\": \"moderate\","
+        "\"preserve_important_words\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/getting-started/basic_usage.md

@@ -0,0 +1,28 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_default();
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/getting-started/extract_file.md

@@ -0,0 +1,39 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("document.pdf", NULL, NULL);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("content:\n%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    char *tables_json = kreuzberg_extraction_result_tables(result);
+    printf("tables (JSON): %s\n", tables_json ? tables_json : "[]");
+    kreuzberg_free_string(tables_json);
+
+    KREUZBERGMetadata *metadata = kreuzberg_extraction_result_metadata(result);
+    if (metadata) {
+        char *title = kreuzberg_metadata_title(metadata);
+        char *language = kreuzberg_metadata_language(metadata);
+        printf("title: %s\n", title ? title : "(none)");
+        printf("language: %s\n", language ? language : "(none)");
+        kreuzberg_free_string(title);
+        kreuzberg_free_string(language);
+        kreuzberg_metadata_free(metadata);
+    }
+
+    kreuzberg_extraction_result_free(result);
+    return 0;
+}
+```

docs/snippets/c/getting-started/extract_with_ocr.md

@@ -0,0 +1,46 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"force_ocr\": true,"
+        "\"ocr\": {\"backend\": \"tesseract\", \"language\": \"eng\"}"
+        "}";
+
+    KREUZBERGExtractionConfig *config =
+        kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config init failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("scanned.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    char *detected_languages = kreuzberg_extraction_result_detected_languages(result);
+    printf("detected languages: %s\n",
+           detected_languages ? detected_languages : "(none)");
+    kreuzberg_free_string(detected_languages);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/getting-started/hello_world.md

@@ -0,0 +1,24 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("document.pdf", NULL, NULL);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    return 0;
+}
+```

docs/snippets/c/getting-started/install_verify.md

@@ -0,0 +1,12 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    const char *version = kreuzberg_version();
+    printf("kreuzberg version: %s\n", version ? version : "(unknown)");
+    return 0;
+}
+```

docs/snippets/c/getting-started/read_content.md

@@ -0,0 +1,38 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file("document.pdf", NULL, NULL);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    if (content) {
+        printf("content length: %zu bytes\n", strlen(content));
+        printf("%s\n", content);
+    }
+    kreuzberg_free_string(content);
+
+    /* Tables are returned as a JSON array string. A real consumer would
+     * feed this into a JSON parser and walk each table's grid. */
+    char *tables_json = kreuzberg_extraction_result_tables(result);
+    if (tables_json) {
+        printf("tables JSON (%zu bytes):\n%s\n",
+               strlen(tables_json), tables_json);
+    } else {
+        printf("tables JSON: (none)\n");
+    }
+    kreuzberg_free_string(tables_json);
+
+    kreuzberg_extraction_result_free(result);
+    return 0;
+}
+```

docs/snippets/c/llm/structured_extraction.md

@@ -0,0 +1,57 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"structured_extraction\": {"
+        "\"schema\": {"
+        "\"type\": \"object\","
+        "\"properties\": {"
+        "\"title\": {\"type\": \"string\"},"
+        "\"authors\": {\"type\": \"array\", \"items\": {\"type\": \"string\"}},"
+        "\"date\": {\"type\": \"string\"}"
+        "},"
+        "\"required\": [\"title\", \"authors\", \"date\"],"
+        "\"additionalProperties\": false"
+        "},"
+        "\"llm\": {\"model\": \"openai/gpt-4o-mini\"},"
+        "\"strict\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("paper.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *structured = kreuzberg_extraction_result_structured_output(result);
+    if (structured) {
+        printf("structured output (JSON):\n%s\n", structured);
+        kreuzberg_free_string(structured);
+    } else {
+        printf("structured output: (none)\n");
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```
+
+<!-- snippet:syntax-only --> Requires network access to the configured LLM provider and a valid API key in the host environment.

docs/snippets/c/mcp/mcp_custom_client.md

@@ -0,0 +1,56 @@
+```c title="C"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+/* The kreuzberg C FFI does not bundle an MCP client. Drive the kreuzberg
+ * CLI's stdio MCP transport from a C host that also links libkreuzberg. */
+int main(void) {
+    int request_pipe[2];
+    int response_pipe[2];
+    if (pipe(request_pipe) < 0 || pipe(response_pipe) < 0) {
+        perror("pipe");
+        return 1;
+    }
+
+    pid_t pid = fork();
+    if (pid < 0) {
+        perror("fork");
+        return 1;
+    }
+    if (pid == 0) {
+        dup2(request_pipe[0], 0);
+        dup2(response_pipe[1], 1);
+        close(request_pipe[1]);
+        close(response_pipe[0]);
+        execlp("kreuzberg", "kreuzberg", "mcp", (char *)NULL);
+        perror("execlp");
+        _exit(127);
+    }
+
+    close(request_pipe[0]);
+    close(response_pipe[1]);
+
+    const char *request =
+        "{\"method\":\"tools/call\","
+        "\"params\":{\"name\":\"extract_file\","
+        "\"arguments\":{\"path\":\"document.pdf\",\"async\":true}}}\n";
+    if (write(request_pipe[1], request, strlen(request)) < 0) {
+        perror("write");
+        return 1;
+    }
+    close(request_pipe[1]);
+
+    char buffer[4096];
+    ssize_t bytes_read = read(response_pipe[0], buffer, sizeof(buffer) - 1);
+    if (bytes_read > 0) {
+        buffer[bytes_read] = '\0';
+        printf("%s", buffer);
+    }
+    close(response_pipe[0]);
+    return 0;
+}
+```
+
+<!-- snippet:syntax-only --> No MCP client is exposed by libkreuzberg; this snippet drives the MCP CLI over stdio.

docs/snippets/c/mcp/mcp_server_start.md

@@ -0,0 +1,29 @@
+```c title="C"
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+/* The kreuzberg C FFI does not embed the MCP server. Spawn the kreuzberg
+ * CLI from a host process that uses libkreuzberg for in-process extraction. */
+int main(void) {
+    pid_t pid = fork();
+    if (pid < 0) {
+        perror("fork");
+        return 1;
+    }
+    if (pid == 0) {
+        execlp("kreuzberg", "kreuzberg", "mcp", (char *)NULL);
+        perror("execlp");
+        _exit(127);
+    }
+
+    int status = 0;
+    if (waitpid(pid, &status, 0) < 0) {
+        perror("waitpid");
+        return 1;
+    }
+    return WIFEXITED(status) ? WEXITSTATUS(status) : 1;
+}
+```
+
+<!-- snippet:syntax-only --> The MCP server is exposed only through the kreuzberg CLI; libkreuzberg's C FFI offers no MCP entry point. This snippet spawns the CLI from a host that already links against libkreuzberg.

docs/snippets/c/metadata/language_detection.md

@@ -0,0 +1,56 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"language_detection\": {"
+        "\"enabled\": true,"
+        "\"min_confidence\": 0.9,"
+        "\"detect_multiple\": false"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    KREUZBERGMetadata *metadata = kreuzberg_extraction_result_metadata(result);
+    if (metadata) {
+        char *language = kreuzberg_metadata_language(metadata);
+        if (language) {
+            printf("Metadata language: %s\n", language);
+            kreuzberg_free_string(language);
+        }
+        kreuzberg_metadata_free(metadata);
+    }
+
+    char *detected_languages_json = kreuzberg_extraction_result_detected_languages(result);
+    if (detected_languages_json) {
+        printf("Detected languages: %s\n", detected_languages_json);
+        kreuzberg_free_string(detected_languages_json);
+    } else {
+        printf("No languages detected\n");
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/metadata/language_detection_multilingual.md

@@ -0,0 +1,46 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"language_detection\": {"
+        "\"enabled\": true,"
+        "\"min_confidence\": 0.8,"
+        "\"detect_multiple\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("multilingual_document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *detected_languages_json = kreuzberg_extraction_result_detected_languages(result);
+    if (detected_languages_json) {
+        printf("Detected languages (JSON array): %s\n", detected_languages_json);
+        kreuzberg_free_string(detected_languages_json);
+    } else {
+        printf("No languages detected\n");
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/metadata/metadata.md

@@ -0,0 +1,27 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    struct CExtractionResult *result = kreuzberg_extract_file_sync("document.pdf");
+    if (!result || !result->success) {
+        fprintf(stderr, "Error: %s\n", kreuzberg_get_error_details().message);
+        return 1;
+    }
+
+    printf("Content: %s\n", result->content);
+    printf("MIME: %s\n", result->mime_type);
+
+    if (result->language)
+        printf("Language: %s\n", result->language);
+    if (result->date)
+        printf("Date: %s\n", result->date);
+    if (result->subject)
+        printf("Subject: %s\n", result->subject);
+    if (result->metadata_json)
+        printf("Metadata: %s\n", result->metadata_json);
+
+    kreuzberg_free_result(result);
+    return 0;
+}
+```

docs/snippets/c/metadata/page_boundaries.md

@@ -0,0 +1,62 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"pages\": {"
+        "\"extract_pages\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    if (content) {
+        printf("Total content length: %zu bytes\n", strlen(content));
+        kreuzberg_free_string(content);
+    }
+
+    KREUZBERGMetadata *metadata = kreuzberg_extraction_result_metadata(result);
+    if (metadata) {
+        KREUZBERGPageStructure *pages = kreuzberg_metadata_pages(metadata);
+        if (pages) {
+            printf("Total pages: %zu\n", kreuzberg_page_structure_total_count(pages));
+
+            char *boundaries_json = kreuzberg_page_structure_boundaries(pages);
+            if (boundaries_json) {
+                printf("Page boundaries (JSON): %s\n", boundaries_json);
+                kreuzberg_free_string(boundaries_json);
+            } else {
+                printf("No page boundaries available\n");
+            }
+            kreuzberg_page_structure_free(pages);
+        } else {
+            printf("No page structure available\n");
+        }
+        kreuzberg_metadata_free(metadata);
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/metadata/page_tracking_basic.md

@@ -0,0 +1,54 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"pages\": {"
+        "\"extract_pages\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *pages_json = kreuzberg_extraction_result_pages(result);
+    if (pages_json) {
+        printf("Pages (JSON array): %s\n", pages_json);
+        kreuzberg_free_string(pages_json);
+    } else {
+        printf("No pages available\n");
+    }
+
+    KREUZBERGMetadata *metadata = kreuzberg_extraction_result_metadata(result);
+    if (metadata) {
+        KREUZBERGPageStructure *pages = kreuzberg_metadata_pages(metadata);
+        if (pages) {
+            printf("Total page count: %zu\n", kreuzberg_page_structure_total_count(pages));
+            kreuzberg_page_structure_free(pages);
+        }
+        kreuzberg_metadata_free(metadata);
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/metadata/tables.md

@@ -0,0 +1,21 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    struct CExtractionResult *result = kreuzberg_extract_file_sync("spreadsheet.xlsx");
+    if (!result || !result->success) {
+        fprintf(stderr, "Error: %s\n", kreuzberg_get_error_details().message);
+        return 1;
+    }
+
+    if (result->tables_json) {
+        printf("Tables (JSON): %s\n", result->tables_json);
+    } else {
+        printf("No tables found\n");
+    }
+
+    kreuzberg_free_result(result);
+    return 0;
+}
+```

docs/snippets/c/metadata/vector_database_integration.md

@@ -0,0 +1,60 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 512,"
+        "\"overlap\": 50,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    if (chunks_json) {
+        printf("Chunks with embeddings (JSON): %s\n", chunks_json);
+        kreuzberg_free_string(chunks_json);
+    } else {
+        printf("No chunks produced\n");
+    }
+
+    KREUZBERGMetadata *metadata = kreuzberg_extraction_result_metadata(result);
+    if (metadata) {
+        char *title = kreuzberg_metadata_title(metadata);
+        if (title) {
+            printf("Document title: %s\n", title);
+            kreuzberg_free_string(title);
+        }
+        kreuzberg_metadata_free(metadata);
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/cloud_ocr_backend.md

@@ -0,0 +1,42 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    /* Cloud OCR backends are registered as custom plugins via the Rust core. */
+    /* Select a registered cloud backend by name through the OCR config. */
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"cloud-ocr\","
+        "\"language\": \"eng\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("scanned.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/image_extraction.md

@@ -0,0 +1,39 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"images\": {"
+        "\"extract_images\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/image_preprocessing.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"images\": {"
+        "\"extract_images\": true,"
+        "\"target_dpi\": 300,"
+        "\"max_image_dimension\": 4096,"
+        "\"auto_adjust_dpi\": true,"
+        "\"min_dpi\": 150,"
+        "\"max_dpi\": 600"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/ocr_easyocr.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"easyocr\","
+        "\"language\": \"en\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/ocr_elements.md

@@ -0,0 +1,43 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"paddleocr\","
+        "\"language\": \"en\","
+        "\"element_config\": {"
+        "\"include_elements\": true"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("scanned.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/ocr_extraction.md

@@ -0,0 +1,26 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    struct ConfigBuilder *builder = kreuzberg_config_builder_new();
+    kreuzberg_config_builder_set_ocr(builder,
+        "{\"tesseract\":{\"language\":\"eng\"}}");
+    ExtractionConfig *config = kreuzberg_config_builder_build(builder);
+
+    char *config_json = kreuzberg_config_to_json(config);
+    struct CExtractionResult *result =
+        kreuzberg_extract_file_sync_with_config("scanned.png", config_json);
+
+    if (result && result->success) {
+        printf("OCR text: %s\n", result->content);
+    } else {
+        fprintf(stderr, "OCR error: %s\n", kreuzberg_get_error_details().message);
+    }
+
+    kreuzberg_free_result(result);
+    kreuzberg_free_string(config_json);
+    kreuzberg_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/ocr_force_all_pages.md

@@ -0,0 +1,41 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"force_ocr\": true,"
+        "\"ocr\": {"
+        "\"backend\": \"tesseract\","
+        "\"language\": \"eng\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/ocr_multi_language.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"tesseract\","
+        "\"language\": \"eng+deu+fra\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("multilingual.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/ocr/ocr_paddleocr.md

@@ -0,0 +1,40 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"ocr\": {"
+        "\"backend\": \"paddleocr\","
+        "\"language\": \"en\""
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/plugins/clear_plugins.md

@@ -0,0 +1,30 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+
+int main(void) {
+    if (kreuzberg_clear_post_processors() != 0) {
+        fprintf(stderr, "clear post-processors failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    if (kreuzberg_clear_ocr_backends() != 0) {
+        fprintf(stderr, "clear OCR backends failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    if (kreuzberg_clear_validators() != 0) {
+        fprintf(stderr, "clear validators failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    printf("All plugins cleared\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/embedding_backend.md

@@ -0,0 +1,115 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* Fixed embedding dimension produced by this backend. */
+#define EMBED_DIM 768
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static uintptr_t dimensions_fn(const void *user_data) {
+    (void)user_data;
+    return (uintptr_t)EMBED_DIM;
+}
+
+static int32_t embed_fn(
+    const void *user_data,
+    const char *texts,
+    char **out_result,
+    char **out_error
+) {
+    (void)user_data;
+    (void)out_error;
+
+    /* `texts` is a JSON array of strings. Count entries by scanning quotes;
+     * a real backend would parse the JSON and call its host model. */
+    size_t count = 0;
+    int in_string = 0;
+    int escape = 0;
+    for (const char *p = texts; *p; ++p) {
+        if (escape) {
+            escape = 0;
+        } else if (*p == '\\') {
+            escape = 1;
+        } else if (*p == '"') {
+            if (!in_string) {
+                in_string = 1;
+                count += 1;
+            } else {
+                in_string = 0;
+            }
+        }
+    }
+
+    /* Build a JSON array of zero vectors of length EMBED_DIM, one per input. */
+    /* Worst case bytes per entry: 2 brackets + EMBED_DIM * 4 ("0.0,") + comma. */
+    size_t cap = 16 + count * (EMBED_DIM * 4 + 4);
+    char *json = (char *)malloc(cap);
+    if (!json) {
+        *out_error = dup_cstr("allocation failure");
+        return 1;
+    }
+    size_t pos = 0;
+    json[pos++] = '[';
+    for (size_t i = 0; i < count; ++i) {
+        if (i > 0) json[pos++] = ',';
+        json[pos++] = '[';
+        for (size_t d = 0; d < EMBED_DIM; ++d) {
+            if (d > 0) json[pos++] = ',';
+            json[pos++] = '0';
+            json[pos++] = '.';
+            json[pos++] = '0';
+        }
+        json[pos++] = ']';
+    }
+    json[pos++] = ']';
+    json[pos] = '\0';
+
+    *out_result = json;
+    return 0;
+}
+
+static void name_fn(const void *user_data, char **out_name) {
+    (void)user_data;
+    *out_name = dup_cstr("my-embedder");
+}
+
+static void version_fn(const void *user_data, char **out_version) {
+    (void)user_data;
+    *out_version = dup_cstr("1.0.0");
+}
+
+int main(void) {
+    KREUZBERGKreuzbergEmbeddingBackendVTable vtable = {0};
+    vtable.name_fn = name_fn;
+    vtable.version_fn = version_fn;
+    vtable.dimensions = dimensions_fn;
+    vtable.embed = embed_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_embedding_backend(
+        "my-embedder",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register embedding backend failed: %s\n",
+                err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("my-embedder registered (dim=%d)\n", EMBED_DIM);
+    return 0;
+}
+```

docs/snippets/c/plugins/extractor_registration.md

@@ -0,0 +1,27 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+
+/*
+ * The kreuzberg C FFI does not expose a public function for registering
+ * custom DocumentExtractor implementations from C. Document extractors must
+ * be registered from Rust via `kreuzberg::plugins::registry::get_document_extractor_registry()`
+ * before the C library is loaded.
+ *
+ * From C you can inspect which extractors the core has registered:
+ */
+
+int main(void) {
+    char *json = kreuzberg_list_document_extractors();
+    if (!json) {
+        fprintf(stderr, "list document extractors failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    printf("Registered document extractors: %s\n", json);
+    kreuzberg_free_string(json);
+    return 0;
+}
+```

docs/snippets/c/plugins/list_plugins.md

@@ -0,0 +1,25 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+
+static void print_plugin_list(const char *label, char *json) {
+    if (!json) {
+        fprintf(stderr, "list %s failed (code %d): %s\n",
+                label,
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return;
+    }
+    printf("%s: %s\n", label, json);
+    kreuzberg_free_string(json);
+}
+
+int main(void) {
+    print_plugin_list("document extractors", kreuzberg_list_document_extractors());
+    print_plugin_list("OCR backends", kreuzberg_list_ocr_backends());
+    print_plugin_list("post-processors", kreuzberg_list_post_processors());
+    print_plugin_list("validators", kreuzberg_list_validators());
+    print_plugin_list("embedding presets", kreuzberg_list_embedding_presets());
+    return 0;
+}
+```

docs/snippets/c/plugins/min_length_validator.md

@@ -0,0 +1,90 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* user_data carries the minimum length threshold. */
+typedef struct {
+    size_t min_length;
+} MinLengthState;
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t validate_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)config;
+    const MinLengthState *state = (const MinLengthState *)user_data;
+
+    /* `result` is a JSON string of ExtractionResult. We approximate the content
+     * length check by scanning for the "content" field. Production plugins
+     * should parse JSON properly. */
+    const char *content = strstr(result, "\"content\":\"");
+    size_t content_len = 0;
+    if (content) {
+        content += strlen("\"content\":\"");
+        const char *end = strchr(content, '"');
+        if (end) {
+            content_len = (size_t)(end - content);
+        }
+    }
+
+    if (content_len < state->min_length) {
+        char buf[128];
+        snprintf(buf, sizeof(buf),
+                 "Content too short: %zu < %zu characters",
+                 content_len,
+                 state->min_length);
+        *out_error = dup_cstr(buf);
+        return 1;
+    }
+    return 0;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 100;
+}
+
+static void free_user_data(void *user_data) {
+    free(user_data);
+}
+
+int main(void) {
+    MinLengthState *state = (MinLengthState *)malloc(sizeof(MinLengthState));
+    state->min_length = 100;
+
+    KREUZBERGKreuzbergValidatorVTable vtable = {0};
+    vtable.validate = validate_fn;
+    vtable.priority = priority_fn;
+    vtable.free_user_data = free_user_data;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_validator(
+        "min-length-validator",
+        vtable,
+        state,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register validator failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        free(state);
+        return 1;
+    }
+
+    printf("min-length-validator registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/pdf_metadata_extractor.md

@@ -0,0 +1,85 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/*
+ * The C FFI does not expose registration for custom DocumentExtractor
+ * implementations. To add PDF-specific behaviour from C, register a
+ * post-processor that runs only on PDF results and enriches them.
+ *
+ * The example below logs whenever the pipeline emits a PDF result, scoped
+ * via the should_process hook so it never fires for other MIME types.
+ */
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)config;
+    (void)out_error;
+    printf("pdf-metadata-extractor: serialised PDF result is %zu bytes\n", strlen(result));
+    return 0;
+}
+
+static int32_t processing_stage_fn(
+    const void *user_data,
+    char **out_result
+) {
+    (void)user_data;
+    *out_result = dup_cstr("\"Late\"");
+    return 0;
+}
+
+static int32_t should_process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config
+) {
+    (void)user_data;
+    (void)config;
+    return strstr(result, "\"mime_type\":\"application/pdf\"") != NULL;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 75;
+}
+
+int main(void) {
+    KREUZBERGKreuzbergPostProcessorVTable vtable = {0};
+    vtable.process = process_fn;
+    vtable.processing_stage = processing_stage_fn;
+    vtable.should_process = should_process_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_post_processor(
+        "pdf-metadata-extractor",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register post-processor failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("pdf-metadata-extractor registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/pdf_only_processor.md

@@ -0,0 +1,78 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)config;
+    (void)out_error;
+
+    printf("pdf-only-processor: handling result of length %zu\n", strlen(result));
+    return 0;
+}
+
+static int32_t processing_stage_fn(
+    const void *user_data,
+    char **out_result
+) {
+    (void)user_data;
+    *out_result = dup_cstr("\"Middle\"");
+    return 0;
+}
+
+static int32_t should_process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config
+) {
+    (void)user_data;
+    (void)config;
+    /* Only process PDF mime types. */
+    return strstr(result, "\"mime_type\":\"application/pdf\"") != NULL;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+int main(void) {
+    KREUZBERGKreuzbergPostProcessorVTable vtable = {0};
+    vtable.process = process_fn;
+    vtable.processing_stage = processing_stage_fn;
+    vtable.should_process = should_process_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_post_processor(
+        "pdf-only-processor",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register post-processor failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("pdf-only-processor registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/plugin_extractor.md

@@ -0,0 +1,42 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <string.h>
+
+/*
+ * The C FFI exposes vtable-based registration for OCR backends, post-processors,
+ * validators, and embedding backends. There is no public C entry point for
+ * registering a custom DocumentExtractor — that must be done from Rust.
+ *
+ * From C you can still drive extraction for any MIME type the Rust core knows
+ * how to handle. The example below feeds JSON bytes through the standard
+ * extraction pipeline by passing the explicit MIME type.
+ */
+
+int main(void) {
+    const char *json_payload = "{\"message\":\"Hello, world!\"}";
+    const uint8_t *bytes = (const uint8_t *)json_payload;
+    uintptr_t bytes_len = (uintptr_t)strlen(json_payload);
+
+    KREUZBERGExtractionResult *result = kreuzberg_extract_bytes_sync(
+        bytes,
+        bytes_len,
+        "application/json",
+        NULL
+    );
+
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("Extracted JSON content: %s\n", content ? content : "(empty)");
+
+    kreuzberg_free_string(content);
+    kreuzberg_extraction_result_free(result);
+    return 0;
+}
+```

docs/snippets/c/plugins/plugin_logging.md

@@ -0,0 +1,91 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* Demonstrates structured logging from a post-processor plugin's lifecycle
+ * hooks (initialize/shutdown) and from the per-result process callback. */
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t initialize_fn(const void *user_data, char **out_error) {
+    (void)user_data;
+    (void)out_error;
+    fprintf(stderr, "[INFO] plugin=logging-demo event=initialize\n");
+    return 0;
+}
+
+static int32_t shutdown_fn(const void *user_data, char **out_error) {
+    (void)user_data;
+    (void)out_error;
+    fprintf(stderr, "[INFO] plugin=logging-demo event=shutdown\n");
+    return 0;
+}
+
+static int32_t process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)config;
+    (void)out_error;
+
+    size_t len = strlen(result);
+    fprintf(stderr,
+            "[INFO] plugin=logging-demo event=process bytes=%zu\n",
+            len);
+
+    if (strstr(result, "\"content\":\"\"") != NULL) {
+        fprintf(stderr,
+                "[WARN] plugin=logging-demo event=empty_content\n");
+    }
+    return 0;
+}
+
+static int32_t processing_stage_fn(const void *user_data, char **out_result) {
+    (void)user_data;
+    *out_result = dup_cstr("\"Late\"");
+    return 0;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+int main(void) {
+    KREUZBERGKreuzbergPostProcessorVTable vtable = {0};
+    vtable.initialize_fn = initialize_fn;
+    vtable.shutdown_fn = shutdown_fn;
+    vtable.process = process_fn;
+    vtable.processing_stage = processing_stage_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_post_processor(
+        "logging-demo",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "[ERROR] register post-processor failed: %s\n",
+                err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("logging-demo post-processor registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/plugin_testing.md

@@ -0,0 +1,84 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* Round-trip test: register a no-op validator, confirm it appears in the
+ * registry list, then unregister and confirm it disappears. */
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t validate_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)result;
+    (void)config;
+    (void)out_error;
+    return 0;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+static int contains_name(const char *json, const char *name) {
+    if (!json || !name) {
+        return 0;
+    }
+    return strstr(json, name) != NULL;
+}
+
+int main(void) {
+    const char *plugin_name = "noop-validator";
+
+    KREUZBERGKreuzbergValidatorVTable vtable = {0};
+    vtable.validate = validate_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    if (kreuzberg_register_validator(plugin_name, vtable, NULL, &err) != 0) {
+        fprintf(stderr, "register failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    char *list_after_register = kreuzberg_list_validators();
+    if (!contains_name(list_after_register, plugin_name)) {
+        fprintf(stderr, "FAIL: validator missing after register\n");
+        kreuzberg_free_string(list_after_register);
+        return 1;
+    }
+    printf("PASS: %s present after register\n", plugin_name);
+    kreuzberg_free_string(list_after_register);
+
+    if (kreuzberg_unregister_validator(plugin_name, &err) != 0) {
+        fprintf(stderr, "unregister failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    char *list_after_unregister = kreuzberg_list_validators();
+    if (contains_name(list_after_unregister, plugin_name)) {
+        fprintf(stderr, "FAIL: validator still present after unregister\n");
+        kreuzberg_free_string(list_after_unregister);
+        return 1;
+    }
+    printf("PASS: %s absent after unregister\n", plugin_name);
+    kreuzberg_free_string(list_after_unregister);
+
+    return 0;
+}
+```

docs/snippets/c/plugins/plugin_validator.md

@@ -0,0 +1,76 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/*
+ * Minimal Validator skeleton: implements the required `validate` function
+ * and the optional `priority` and `should_validate` hooks via the C vtable.
+ */
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t validate_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)config;
+
+    /* Reject results whose serialised form contains a clearly forbidden token. */
+    if (strstr(result, "FORBIDDEN") != NULL) {
+        *out_error = dup_cstr("Content contains forbidden token");
+        return 1;
+    }
+    return 0;
+}
+
+static int32_t should_validate_fn(
+    const void *user_data,
+    const char *result,
+    const char *config
+) {
+    (void)user_data;
+    (void)result;
+    (void)config;
+    return 1;  /* always run */
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+int main(void) {
+    KREUZBERGKreuzbergValidatorVTable vtable = {0};
+    vtable.validate = validate_fn;
+    vtable.should_validate = should_validate_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_validator(
+        "forbidden-token-validator",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register validator failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("forbidden-token-validator registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/quality_score_validator.md

@@ -0,0 +1,70 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t validate_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)config;
+
+    /* Look for a "quality_score" key inside the metadata.additional map.
+     * Production plugins should parse the JSON properly. */
+    double score = 0.0;
+    const char *needle = "\"quality_score\":";
+    const char *found = strstr(result, needle);
+    if (found) {
+        score = atof(found + strlen(needle));
+    }
+
+    if (score < 0.5) {
+        char buf[128];
+        snprintf(buf, sizeof(buf),
+                 "Quality score too low: %.2f < 0.50", score);
+        *out_error = dup_cstr(buf);
+        return 1;
+    }
+    return 0;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+int main(void) {
+    KREUZBERGKreuzbergValidatorVTable vtable = {0};
+    vtable.validate = validate_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_validator(
+        "quality-score-validator",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register validator failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("quality-score-validator registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/stateful_plugin.md

@@ -0,0 +1,101 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdatomic.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* Shared state lives in `user_data` and is forwarded to every vtable callback.
+ * Use atomics or a mutex if more than one thread can call into the plugin. */
+
+typedef struct {
+    atomic_size_t call_count;
+} StatefulState;
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t initialize_fn(const void *user_data, char **out_error) {
+    (void)out_error;
+    StatefulState *state = (StatefulState *)user_data;
+    atomic_store(&state->call_count, 0);
+    return 0;
+}
+
+static int32_t shutdown_fn(const void *user_data, char **out_error) {
+    (void)out_error;
+    const StatefulState *state = (const StatefulState *)user_data;
+    size_t count = atomic_load(&state->call_count);
+    fprintf(stderr, "stateful-plugin: shutdown after %zu calls\n", count);
+    return 0;
+}
+
+static int32_t process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)result;
+    (void)config;
+    (void)out_error;
+    StatefulState *state = (StatefulState *)user_data;
+    atomic_fetch_add(&state->call_count, 1);
+    return 0;
+}
+
+static int32_t processing_stage_fn(const void *user_data, char **out_result) {
+    (void)user_data;
+    *out_result = dup_cstr("\"Middle\"");
+    return 0;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+static void free_user_data(void *user_data) {
+    free(user_data);
+}
+
+int main(void) {
+    StatefulState *state = (StatefulState *)malloc(sizeof(StatefulState));
+    if (!state) {
+        return 1;
+    }
+    atomic_init(&state->call_count, 0);
+
+    KREUZBERGKreuzbergPostProcessorVTable vtable = {0};
+    vtable.initialize_fn = initialize_fn;
+    vtable.shutdown_fn = shutdown_fn;
+    vtable.process = process_fn;
+    vtable.processing_stage = processing_stage_fn;
+    vtable.priority = priority_fn;
+    vtable.free_user_data = free_user_data;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_post_processor(
+        "stateful-plugin",
+        vtable,
+        state,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register post-processor failed: %s\n",
+                err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        free(state);
+        return 1;
+    }
+
+    printf("stateful-plugin registered\n");
+    return 0;
+}
+```

docs/snippets/c/plugins/unregister_plugins.md

@@ -0,0 +1,31 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+
+static int unregister_or_log(
+    int32_t (*unregister_fn)(const char *, char **),
+    const char *kind,
+    const char *name
+) {
+    char *err = NULL;
+    int32_t rc = unregister_fn(name, &err);
+    if (rc != 0) {
+        fprintf(stderr, "unregister %s '%s' failed: %s\n",
+                kind,
+                name,
+                err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+    return 0;
+}
+
+int main(void) {
+    int failures = 0;
+    failures += unregister_or_log(kreuzberg_unregister_post_processor, "post-processor", "word-count");
+    failures += unregister_or_log(kreuzberg_unregister_validator, "validator", "min-length-validator");
+    failures += unregister_or_log(kreuzberg_unregister_ocr_backend, "OCR backend", "my-ocr");
+    failures += unregister_or_log(kreuzberg_unregister_embedding_backend, "embedding backend", "my-embedder");
+    return failures == 0 ? 0 : 1;
+}
+```

docs/snippets/c/plugins/word_count_processor.md

@@ -0,0 +1,92 @@
+```c title="C"
+#include <kreuzberg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+static char *dup_cstr(const char *s) {
+    size_t len = strlen(s);
+    char *out = (char *)malloc(len + 1);
+    if (out) {
+        memcpy(out, s, len + 1);
+    }
+    return out;
+}
+
+static int32_t process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config,
+    char **out_error
+) {
+    (void)user_data;
+    (void)config;
+    (void)out_error;
+
+    /* The `result` JSON string is read-only at this layer; for a real
+     * mutating post-processor, decode the JSON, mutate, and serialise back
+     * via the kreuzberg ExtractionResult helpers in your host language. */
+    size_t words = 0;
+    int in_word = 0;
+    for (const char *p = result; *p; ++p) {
+        if (*p == ' ' || *p == '\t' || *p == '\n' || *p == '\r') {
+            in_word = 0;
+        } else if (!in_word) {
+            in_word = 1;
+            words += 1;
+        }
+    }
+    printf("word-count: ~%zu tokens in serialised result\n", words);
+    return 0;
+}
+
+static int32_t processing_stage_fn(
+    const void *user_data,
+    char **out_result
+) {
+    (void)user_data;
+    /* ProcessingStage is JSON-serialised; "Early" maps to ProcessingStage::Early. */
+    *out_result = dup_cstr("\"Early\"");
+    return 0;
+}
+
+static int32_t should_process_fn(
+    const void *user_data,
+    const char *result,
+    const char *config
+) {
+    (void)user_data;
+    (void)config;
+    /* Skip empty content. */
+    return strstr(result, "\"content\":\"\"") == NULL;
+}
+
+static int32_t priority_fn(const void *user_data) {
+    (void)user_data;
+    return 50;
+}
+
+int main(void) {
+    KREUZBERGKreuzbergPostProcessorVTable vtable = {0};
+    vtable.process = process_fn;
+    vtable.processing_stage = processing_stage_fn;
+    vtable.should_process = should_process_fn;
+    vtable.priority = priority_fn;
+
+    char *err = NULL;
+    int32_t rc = kreuzberg_register_post_processor(
+        "word-count",
+        vtable,
+        NULL,
+        &err
+    );
+    if (rc != 0) {
+        fprintf(stderr, "register post-processor failed: %s\n", err ? err : "(no detail)");
+        kreuzberg_free_string(err);
+        return 1;
+    }
+
+    printf("word-count post-processor registered\n");
+    return 0;
+}
+```

docs/snippets/c/utils/chunking.md

@@ -0,0 +1,41 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"max_characters\": 1500,"
+        "\"overlap\": 200"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks (JSON): %s\n", chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/chunking_rag.md

@@ -0,0 +1,49 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 500,"
+        "\"overlap\": 50,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("research_paper.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    /* Each chunk JSON entry contains content, embedding, and metadata
+       (chunk_index, total_chunks, byte_start, byte_end). Pipe this directly
+       into a vector database client. */
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks (JSON):\n%s\n", chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/embedding_with_chunking.md

@@ -0,0 +1,48 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 1024,"
+        "\"overlap\": 100,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true,"
+        "\"batch_size\": 32,"
+        "\"show_download_progress\": false"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("chunks with embeddings (JSON):\n%s\n", chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/keyword_extraction_example.md

@@ -0,0 +1,46 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"keywords\": {"
+        "\"algorithm\": \"yake\","
+        "\"max_keywords\": 10,"
+        "\"min_score\": 0.3"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("research_paper.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *keywords_json = kreuzberg_extraction_result_extracted_keywords(result);
+    if (keywords_json) {
+        printf("Keywords: %s\n", keywords_json);
+        kreuzberg_free_string(keywords_json);
+    } else {
+        printf("Keywords: (none)\n");
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/quality_processing_example.md

@@ -0,0 +1,45 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"enable_quality_processing\": true"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("scanned_document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    double score = kreuzberg_extraction_result_quality_score(result);
+    if (score < 0.5) {
+        printf("Warning: low quality extraction (%.2f)\n", score);
+    } else {
+        printf("Quality score: %.2f\n", score);
+    }
+
+    char *warnings_json = kreuzberg_extraction_result_processing_warnings(result);
+    printf("processing warnings (JSON): %s\n", warnings_json ? warnings_json : "[]");
+    kreuzberg_free_string(warnings_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/standalone_embed.md

@@ -0,0 +1,39 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true"
+        "}";
+
+    KREUZBERGEmbeddingConfig *config = kreuzberg_embedding_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    /* Embed input is a JSON-encoded array of strings. */
+    const char *texts_json = "[\"Hello, world!\", \"Kreuzberg is fast\"]";
+
+    char *embeddings_json = kreuzberg_embed_texts(texts_json, config);
+    if (!embeddings_json) {
+        fprintf(stderr, "embedding failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_embedding_config_free(config);
+        return 1;
+    }
+
+    printf("embeddings (JSON, 2D float array):\n%s\n", embeddings_json);
+    kreuzberg_free_string(embeddings_json);
+
+    kreuzberg_embedding_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/token_reduction.md

@@ -0,0 +1,41 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"token_reduction\": {"
+        "\"mode\": \"moderate\","
+        "\"preserve_important_words\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    printf("reduced content:\n%s\n", content ? content : "(empty)");
+    kreuzberg_free_string(content);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/token_reduction_example.md

@@ -0,0 +1,44 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+    const char *config_json =
+        "{"
+        "\"token_reduction\": {"
+        "\"mode\": \"moderate\","
+        "\"preserve_important_words\": true"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync("verbose_document.pdf", NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    char *content = kreuzberg_extraction_result_content(result);
+    if (content) {
+        printf("reduced content (%zu bytes):\n%s\n", strlen(content), content);
+        kreuzberg_free_string(content);
+    }
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/c/utils/vector_database_integration.md

@@ -0,0 +1,55 @@
+```c title="C"
+#include "kreuzberg.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(void) {
+    const char *document_path = "document.pdf";
+    const char *document_id = "doc-001";
+
+    const char *config_json =
+        "{"
+        "\"chunking\": {"
+        "\"chunker_type\": \"character\","
+        "\"max_characters\": 512,"
+        "\"overlap\": 50,"
+        "\"embedding\": {"
+        "\"model\": {\"preset\": {\"name\": \"balanced\"}},"
+        "\"normalize\": true,"
+        "\"batch_size\": 32"
+        "}"
+        "}"
+        "}";
+
+    KREUZBERGExtractionConfig *config = kreuzberg_extraction_config_from_json(config_json);
+    if (!config) {
+        fprintf(stderr, "config parse failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        return 1;
+    }
+
+    KREUZBERGExtractionResult *result =
+        kreuzberg_extract_file_sync(document_path, NULL, config);
+    if (!result) {
+        fprintf(stderr, "extraction failed (code %d): %s\n",
+                kreuzberg_last_error_code(),
+                kreuzberg_last_error_context());
+        kreuzberg_extraction_config_free(config);
+        return 1;
+    }
+
+    /* The chunks JSON array carries content + embedding + metadata for each
+       chunk. Pass this directly to your vector database client (pgvector,
+       Qdrant, Pinecone, etc.) along with the document_id as a metadata field. */
+    char *chunks_json = kreuzberg_extraction_result_chunks(result);
+    printf("document_id: %s\n", document_id);
+    printf("chunks (JSON, ready to upsert into a vector DB):\n%s\n",
+           chunks_json ? chunks_json : "[]");
+    kreuzberg_free_string(chunks_json);
+
+    kreuzberg_extraction_result_free(result);
+    kreuzberg_extraction_config_free(config);
+    return 0;
+}
+```

docs/snippets/cli/batch_basic.md

@@ -0,0 +1,7 @@
+```bash title="Bash"
+# Process multiple files
+kreuzberg extract doc1.pdf doc2.docx doc3.pptx
+
+# Use glob patterns
+kreuzberg extract documents/**/*.pdf
+```