LongTrainer 1.3.1 — Production-Ready RAG Framework

Multi-tenant bots, streaming, tools, and persistent memory — all batteries included.

---

Welcome to LongTrainer 1.3.1

LongTrainer is a production-ready RAG framework that turns your documents into intelligent, multi-tenant chatbots with minimal code. Built on top of LangChain, it handles multi-bot isolation, persistent MongoDB memory, FAISS vector search, streaming responses, custom tool calling, chat encryption, and vision support.

Quick Start

Install LongTrainer and start building in minutes:

pip install longtrainer

RAG Mode (Default)

from longtrainer.trainer import LongTrainer
import os

os.environ["OPENAI_API_KEY"] = "sk-..."

trainer = LongTrainer(mongo_endpoint="mongodb://localhost:27017/")
bot_id = trainer.initialize_bot_id()

trainer.add_document_from_path("data.pdf", bot_id)
trainer.create_bot(bot_id)

chat_id = trainer.new_chat(bot_id)
answer, sources = trainer.get_response("What is this about?", bot_id, chat_id)
print(answer)

Agent Mode (With Tools)

from longtrainer.tools import web_search
from langchain_core.tools import tool

@tool
def calculate(expression: str) -> str:
    """Evaluate a math expression."""
    return str(eval(expression))

trainer.add_tool(web_search, bot_id)
trainer.add_tool(calculate, bot_id)

trainer.create_bot(bot_id, agent_mode=True)
chat_id = trainer.new_chat(bot_id)
answer, _ = trainer.get_response("What is 42 * 17?", bot_id, chat_id)

Zero-Code CLI & API Server

Manage bots, chat, and run a production API directly from your terminal—no Python required.

A. Interactive Terminal Chat

# 1. Initialize a new project and generate longtrainer.yaml
longtrainer init

# 2. Create a new bot
longtrainer bot create --prompt "You are a helpful assistant."

# 3. Add a document (PDF, link, etc.)
longtrainer add-doc <bot_id> /path/to/document.pdf

# 4. Start chatting!
longtrainer chat <bot_id>

B. Interactive Streaming Chat

longtrainer chat <bot_id>

C. FastAPI REST Server

Start a production-ready API server backed by your LongTrainer bots:

longtrainer serve

Visit http://localhost:8000/docs to see the auto-generated Swagger UI.

What's New in 1.2.2

• Dynamic Tool Engine: Inject any LangChain tool by string name — zero custom code ("wikipedia", "arxiv", "tavily_search_results_json")
• Enterprise Vector DBs: 9 providers — FAISS, Pinecone, Chroma, Qdrant, PGVector, MongoDB Atlas, Milvus, Weaviate, Elasticsearch
• Enterprise Document Loaders: AWS S3, Google Drive, Confluence, GitHub, JSON, Notion, and dynamic loader injection
• Dynamic Model Factory: OpenAI, Anthropic, Google, AWS Bedrock, HuggingFace, Groq, Together, Ollama
• CLI & API Enhancements: --tools flag and API tools parameter for zero-code tool injection
• Tool Persistence: Tool selections saved to MongoDB and restored on bot reload

What's New in 1.3.1

• LongTracer Integration: Native, optional observability and hallucination detection for all response paths (RAG, Agent, Vision, Structured)
• CitationVerifier: Hybrid STS + NLI claim verification to detect hallucinated responses
• Lightweight Tracing: tracer_verify=False for spans-only observability without the ~500MB NLI model download
• Graceful Degradation: Zero overhead when disabled, no crash when longtracer is not installed
• Lazy Loading: Chat histories are lazy-loaded on demand from MongoDB, eliminating startup latency

What's New in 1.0.0

• Dual Mode: RAG (LCEL) for simple Q&A, Agent (LangGraph) for tool calling
• Streaming Responses: Sync and async out of the box
• Custom Tool Calling: add_tool() with any LangChain @tool
• Per-Bot Customization: Independent LLM, embeddings, and retrieval config per bot
• Chat Encryption: Fernet encryption for stored conversations

Upgrading from 0.3.4? See the Migration Guide.

Documentation

Guide	Description
Installation	Install LongTrainer and system dependencies
Creating an Instance	Configure the LongTrainer class
Creating and Using a Bot	Bot lifecycle: create, load, chat
Agent Mode & Tools	Tool calling, streaming, agent configuration
CLI & API Server	Zero-code CLI and REST API
Rate Limiting	Two-layer execution rate limit configuration
Chat Management	Sessions, history, training on chats
Supported Formats	Document types and ingestion methods
Integrating LLMs	Use any LangChain-compatible LLM
Integrating Embeddings	Custom embedding models
Observability & Tracing	LongTracer hallucination detection & spans
Updating Bots	Add documents and reconfigure bots
Deleting Bots	Remove bots and associated data
Migration 0.3.4 → 1.0.0	Breaking changes and upgrade path

Part of the Long Suite

LongTrainer is part of the Long Suite — a collection of tools for building, testing, and monitoring production RAG systems.

Project	Description
LongParser	Document ingestion and chunking
LongTrainer	RAG chatbot framework (you are here)
LongTracer	Hallucination detection and tracing
LongProbe	Retrieval regression testing