Files
osmedeus/docs/api/llm.mdx
T

160 lines
3.9 KiB
Plaintext

---
title: "LLM API"
description: "OpenAI-compatible chat completions and embeddings"
---
# LLM API
Direct API access to Large Language Model capabilities without requiring workflow execution.
## Chat Completion
Send a chat completion request to the configured LLM provider.
```bash
curl -X POST http://localhost:8002/osm/api/llm/v1/chat/completions \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{
"messages": [
{"role": "system", "content": "You are a security analyst."},
{"role": "user", "content": "Analyze the security posture of example.com"}
],
"max_tokens": 1000,
"temperature": 0.7
}'
```
**Request Body:**
| Field | Type | Required | Description |
|-------|------|----------|-------------|
| `messages` | array | Yes | Array of message objects with `role` and `content` |
| `model` | string | No | Model to use (defaults to provider's default model) |
| `max_tokens` | int | No | Maximum tokens in response |
| `temperature` | float | No | Sampling temperature (0.0-2.0) |
| `top_p` | float | No | Top-p sampling parameter |
| `top_k` | int | No | Top-k sampling parameter |
| `n` | int | No | Number of completions to generate |
| `stream` | bool | No | Enable streaming (not yet supported) |
| `tools` | array | No | Tool definitions for function calling |
| `tool_choice` | string/object | No | Tool selection strategy |
| `response_format` | object | No | Response format (`{"type": "json_object"}`) |
**Message Roles:**
- `system` - System prompt to set assistant behavior
- `user` - User message
- `assistant` - Previous assistant response
- `tool` - Tool call result
**Response:**
```json
{
"id": "chatcmpl-abc123",
"model": "gpt-4",
"content": "Based on my analysis of example.com...",
"finish_reason": "stop",
"usage": {
"prompt_tokens": 50,
"completion_tokens": 200,
"total_tokens": 250
}
}
```
---
## With Tools (Function Calling)
```bash
curl -X POST http://localhost:8002/osm/api/llm/v1/chat/completions \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{
"messages": [
{"role": "user", "content": "What DNS records exist for example.com?"}
],
"tools": [
{
"type": "function",
"function": {
"name": "dns_lookup",
"description": "Look up DNS records for a domain",
"parameters": {
"type": "object",
"properties": {
"domain": {"type": "string", "description": "Domain to look up"},
"record_type": {"type": "string", "enum": ["A", "AAAA", "MX", "TXT", "NS"]}
},
"required": ["domain"]
}
}
}
],
"tool_choice": "auto"
}'
```
**Response with Tool Calls:**
```json
{
"id": "chatcmpl-xyz789",
"model": "gpt-4",
"content": null,
"finish_reason": "tool_calls",
"tool_calls": [
{
"id": "call_abc123",
"type": "function",
"function": {
"name": "dns_lookup",
"arguments": "{\"domain\": \"example.com\", \"record_type\": \"A\"}"
}
}
],
"usage": {
"prompt_tokens": 100,
"completion_tokens": 25,
"total_tokens": 125
}
}
```
---
## Generate Embeddings
Generate vector embeddings for input text.
```bash
curl -X POST http://localhost:8002/osm/api/llm/v1/embeddings \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{
"input": ["security analysis", "vulnerability assessment"],
"model": "text-embedding-3-small"
}'
```
**Request Body:**
| Field | Type | Required | Description |
|-------|------|----------|-------------|
| `input` | array | Yes | Array of strings to embed |
| `model` | string | No | Embedding model (defaults to provider's model) |
**Response:**
```json
{
"model": "text-embedding-3-small",
"embeddings": [
[0.0023, -0.0045, 0.0178, ...],
[0.0112, -0.0067, 0.0234, ...]
],
"usage": {
"prompt_tokens": 10,
"total_tokens": 10
}
}
```