ai_ollama.c File Reference

Ollama driver for the AI unified interface. More...

#include <curl/curl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "cJSON.h"
#include "csilk/drivers/ai.h"

Include dependency graph for ai_ollama.c:

Data Structures
struct	ollama_state_t
	Per-instance state for the Ollama driver. More...
struct	curl_response
	Accumulates a complete HTTP response body in memory. More...

Functions
static void *	ollama_init (const char *api_key, const char *base_url)
	Initialize the Ollama driver state.
static void	ollama_free (void *state_ptr)
	Free the Ollama driver state and associated resources.
static size_t	write_cb (void *contents, size_t size, size_t nmemb, void *userp)
	libcurl write callback that appends data to a growing buffer.
static int	ollama_chat (void *state_ptr, const csilk_ai_chat_request_t *req, csilk_ai_chat_response_t *res)
	Execute a chat completion against the Ollama /api/chat endpoint.
static int	ollama_embeddings (void *state_ptr, const char *model, const char **input, size_t count, csilk_ai_embeddings_response_t *res)
	Generate embeddings via Ollama (stub – not yet implemented).
void	csilk_ai_ollama_init_driver (void)
	Register the Ollama driver with the AI subsystem. Called during startup to make "ollama" available to csilk_ai_new().

Variables
static const csilk_ai_driver_t	ollama_driver
	Driver vtable for the Ollama AI backend.

Detailed Description

Ollama driver for the AI unified interface.

Implements the csilk_ai_driver_t vtable for local LLM inference via Ollama. Key differences from the OpenAI driver:

• No API key required (Ollama runs locally by default).
• Chat uses POST /api/chat with a nested "options" object for parameters.
• Token counts use "prompt_eval_count" / "eval_count" instead of "usage".
• No streaming support yet (simplified with "stream": false).

Copyright

MIT License

---

Data Structure Documentation

ollama_state_t

struct ollama_state_t

Per-instance state for the Ollama driver.

Data Fields
char *	base_url	Base URL of the Ollama server (e.g., http://localhost:11434).

curl_response

struct curl_response

Accumulates a complete HTTP response body in memory.

Accumulates a complete HTTP response body (used for embeddings).

Data Fields
char *	body	Heap-allocated body buffer (grown on each write callback). Heap-allocated body buffer.
size_t	size	Current length of data in body (excluding NUL terminator). Current length of data in body.

Function Documentation

csilk_ai_ollama_init_driver()

void csilk_ai_ollama_init_driver

(

void

)

Register the Ollama driver with the AI subsystem. Called during startup to make "ollama" available to csilk_ai_new().

ollama_chat()

static int ollama_chat ( void *  state_ptr, const csilk_ai_chat_request_t *  req, csilk_ai_chat_response_t *  res  )

static

Execute a chat completion against the Ollama /api/chat endpoint.

Flow:

1. Build a JSON request body with messages array + nested "options". Ollama differs from OpenAI by placing sampling params under "options": { "temperature": ..., "top_p": ... } instead of at the root.
2. POST the JSON to <base_url>/api/chat with Content-Type: application/json.
3. On HTTP success, parse the response extracting "message"."content" for the reply text and "prompt_eval_count" / "eval_count" for token usage.
4. Returns 0 on success, -1 on transport or parse failure.

Note

Streaming is not yet implemented (always sends "stream": false).

No auth headers are set – Ollama's local server does not require one.

ollama_embeddings()

static int ollama_embeddings ( void *  state_ptr, const char *  model, const char **  input, size_t  count, csilk_ai_embeddings_response_t *  res  )

static

Generate embeddings via Ollama (stub – not yet implemented).

Todo:

Implement using Ollama's /api/embeddings endpoint.

ollama_free()

static void ollama_free ( void *  state_ptr )

static

Free the Ollama driver state and associated resources.

ollama_init()

static void * ollama_init ( const char *  api_key, const char *  base_url  )

static

Initialize the Ollama driver state.

Note

api_key is accepted but ignored – Ollama's local server requires no authentication. Allocates state, sets the server URL (defaulting to localhost:11434), and performs one-shot libcurl global init.

write_cb()

static size_t write_cb ( void *  contents, size_t  size, size_t  nmemb, void *  userp  )

static

libcurl write callback that appends data to a growing buffer.

Returns

Number of bytes "consumed" (must equal realsize for the transfer to continue), or 0 to abort the transfer on OOM.

Variable Documentation

ollama_driver

const csilk_ai_driver_t ollama_driver

static

Initial value:

= {
    .name = "ollama",
    .init = ollama_init,
    .chat = ollama_chat,
    .embeddings = ollama_embeddings,
    .free = ollama_free,
}

Driver vtable for the Ollama AI backend.