Nomad changes
All checks were successful
Deploy fil (kreuzberg) / deploy (push) Successful in 49s

This commit is contained in:
Henrik Jess Nielsen
2026-06-01 23:40:55 +02:00
parent 72b1a0a6ed
commit b4c07d3693
5723 changed files with 1130655 additions and 0 deletions

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```

View File

@@ -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;
}
```