API Integration

LlamaFarm API Integration Skill

Guide for using LlamaFarm's REST API programmatically.

When to Load

Load this skill when the user:

• Wants to integrate LlamaFarm into an application
• Needs to make API calls programmatically
• Is building automation scripts
• Asks about HTTP endpoints or SDK usage
• Wants to use LlamaFarm from Python/JavaScript/curl

API Overview

LlamaFarm exposes an OpenAI-compatible REST API at http://localhost:14345/v1.

Quick Start

Check Server Health

curl http://localhost:14345/health

Chat with RAG

curl -X POST http://localhost:14345/v1/projects/default/my-project/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "What is in my documents?"}],
    "rag": true
  }'

Query Vector Database

curl -X POST http://localhost:14345/v1/projects/default/my-project/rag/query \
  -H "Content-Type: application/json" \
  -d '{
    "query": "neural scaling laws",
    "database": "main_db",
    "top_k": 5
  }'

OpenAI Compatibility

The chat endpoint is compatible with OpenAI's API format:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:14345/v1/projects/default/my-project",
    api_key="not-needed"  # LlamaFarm doesn't require API key by default
)

response = client.chat.completions.create(
    model="default",
    messages=[
        {"role": "user", "content": "Hello!"}
    ]
)
print(response.choices[0].message.content)

Progressive Disclosure

For detailed endpoint documentation:

• rest-api.md - Complete endpoint reference
• chat-completions.md - Chat API details
• datasets-api.md - Dataset management
• rag-api.md - RAG query operations

Common Integration Patterns

1. Simple Chat Application

import requests

def chat(message, project="my-project", namespace="default"):
    response = requests.post(
        f"http://localhost:14345/v1/projects/{namespace}/{project}/chat/completions",
        json={
            "messages": [{"role": "user", "content": message}],
            "rag": True
        }
    )
    return response.json()["choices"][0]["message"]["content"]

answer = chat("What are the main topics in my documents?")

2. Batch Document Upload

import requests
from pathlib import Path

def upload_documents(files, dataset, project="my-project"):
    for file_path in files:
        with open(file_path, "rb") as f:
            requests.post(
                f"http://localhost:14345/v1/projects/default/{project}/datasets/{dataset}/data",
                files={"file": f}
            )

3. Streaming Responses

import requests

def stream_chat(message, project="my-project"):
    response = requests.post(
        f"http://localhost:14345/v1/projects/default/{project}/chat/completions",
        json={
            "messages": [{"role": "user", "content": message}],
            "stream": True
        },
        stream=True
    )
    for line in response.iter_lines():
        if line:
            print(line.decode())

Authentication

By default, LlamaFarm doesn't require authentication for local access.

For production deployments, configure authentication:

# In llamafarm.yaml
server:
  auth:
    enabled: true
    api_key: ${env:LLAMAFARM_API_KEY}

Then include in requests:

curl -H "Authorization: Bearer your-api-key" ...

Error Handling

All endpoints return standard error responses:

{
  "error": {
    "code": "not_found",
    "message": "Project 'unknown' not found in namespace 'default'"
  }
}

Common HTTP status codes: