Add the LangChain Advanced RAG, MCP, and Mastra partner cookbooks

partners/README.md

@@ -12,7 +12,10 @@ Each partner brings their unique expertise and technology stack to create compre
 | **Galileo** | AI Hallucination Detection and Reduction | AI Safety - Techniques for detecting and reducing hallucinations in AI systems to improve reliability and accuracy | [View Notebook](galileo/ai_hallucination_detection_and_reduction.ipynb) |
 | **Gravity9** | Agentic System Enhanced Contract and Supply Chain Management for International Shipping | Supply Chain Management - Agentic RAG solution for contract and supply chain management in international shipping | [View Notebook](gravity9/Agentic_System_Enhanced_Contract_and_Supply_Chain_Management_for_International_Shipping.ipynb) |
 | **LangChain** | Agentic Knowledge Discovery | Knowledge Discovery - Advanced agentic systems for intelligent knowledge discovery and information retrieval | [View Notebook](langchain/agentic_knowledge_discovery_notebook.ipynb) |
+| **LangChain** | Advanced RAG (3 notebooks) | Movie Discovery - Vector, full-text, hybrid `$rankFusion`, reranking, `$graphLookup`, and auto-embedding retrieval with LangChain on a free M0 cluster | [View Notebooks](langchain/advanced-rag/) |
+| **LangChain** | MongoDB MCP Server + AI Agents | Database Agents - A LangGraph agent that answers questions about a cluster by calling MongoDB MCP Server tools | [View Notebook](langchain/claude-code-mcp/01_mongodb_mcp_and_agent.ipynb) |
 | **LlamaIndex** | Cemex + MongoDB + LlamaCloud Demo | Enterprise Integration - Demonstration of MongoDB integration with LlamaCloud using real-world enterprise data from Cemex | [View Notebook](llamaindex/Cemex+MongoDB+LlamaCloud_demo.ipynb) |
+| **Mastra** | Movie-Discovery Agent with Mastra (4 notebooks, TypeScript) | Agents - Vector, hybrid, and reranked retrieval, bring-your-own collections, tools, and persistent memory with Mastra's MongoDB integration | [View Notebooks](mastra/) |
 
 ## About Our Partners
 
@@ -31,6 +34,9 @@ Each partner brings their unique expertise and technology stack to create compre
 ### LlamaIndex
 [LlamaIndex](https://www.llamaindex.ai/) is a data framework for building LLM applications, providing tools for data ingestion, indexing, and querying to connect custom data sources with large language models.
 
+### Mastra
+[Mastra](https://mastra.ai/) is a TypeScript framework for building AI agents and workflows, with built-in memory, tools, and retrieval backed by stores such as MongoDB.
+
 ---
 
 These partnerships demonstrate MongoDB's commitment to fostering innovation in the AI ecosystem and providing developers with comprehensive solutions for building intelligent applications across various domains and industries.

partners/langchain/.env.example

@@ -0,0 +1,33 @@
+# Copy to .env in this folder and fill it in. Both sets here, advanced-rag/ and
+# claude-code-mcp/, read it via python-dotenv, which finds the nearest .env at or above
+# the notebook's folder. A missing value falls back to an interactive getpass prompt, so
+# Colab works without a .env at all.
+
+# --- MongoDB Atlas ---
+# Your Atlas M0 (free tier) SRV connection string.
+MONGODB_URI=mongodb+srv://<user>:<password>@<cluster>.mongodb.net/?retryWrites=true&w=majority
+MONGODB_DB=sample_mflix
+# Advanced RAG notebook 3 creates one scratch collection for its auto-embedding demo and
+# drops it during cleanup. It refuses to reuse a collection that already exists, so set
+# this only if the default name is already taken on your cluster.
+LANGCHAIN_AUTO_EMBED_COLLECTION=auto_embed_demo
+
+# --- Atlas-hosted embeddings + reranking (ai.mongodb.com) ---
+# Atlas Admin API key used to call the hosted embedding and rerank endpoints.
+MDB_ATLAS_API_KEY=
+# sample_mflix.embedded_movies ships with voyage-3-large (2048-d) plot vectors,
+# so query embeddings use the same model/dimensions to share the vector space.
+EMBEDDING_MODEL=voyage-3-large
+EMBEDDING_DIMENSIONS=2048
+
+# --- Claude (LLM) ---
+ANTHROPIC_API_KEY=
+ANTHROPIC_MODEL=claude-haiku-4-5
+# Optional: point ChatAnthropic at an Anthropic-compatible gateway (e.g. Grove).
+# Leave blank to use the default Anthropic API (api.anthropic.com).
+# Give the ROOT url WITHOUT the trailing /v1/messages -- the SDK appends that
+# itself, e.g. https://<gateway-host>/<path>/anthropic
+ANTHROPIC_BASE_URL=
+# Some gateways (e.g. Azure API Management fronting Grove) authenticate with an
+# "api-key" header instead of Anthropic's default "x-api-key". When ANTHROPIC_BASE_URL
+# is set, the notebooks send ANTHROPIC_API_KEY in the "api-key" header automatically.

partners/langchain/.gitignore

@@ -0,0 +1,2 @@
+__pycache__/
+.pytest_cache/

partners/langchain/README.md

@@ -0,0 +1,58 @@
+# LangChain + MongoDB Atlas
+
+Two notebook sets built on LangChain and LangGraph, in Python. Both run on a free Atlas M0
+cluster against the `sample_mflix` sample dataset, and both open in Colab.
+
+| Set | What you'll build | Notebooks |
+|-----|-------------------|-----------|
+| [Advanced RAG](./advanced-rag/) | A movie-discovery assistant that retrieves with vector, full-text, hybrid, graph, and auto-embedding techniques | 3 |
+| [Claude Code & MCP](./claude-code-mcp/) | An agent that answers questions about your cluster by calling MongoDB MCP Server tools | 1 |
+
+The MCP set lives here because its agent is built with `langgraph` and
+`langchain-mcp-adapters`. MongoDB's MCP Server is the subject; LangChain is how the agent
+around it is assembled.
+
+## What you need
+
+- A free Atlas **M0** cluster with **Load Sample Dataset** run (see
+  [Load Sample Data](https://www.mongodb.com/docs/atlas/sample-data/)).
+- An **Atlas API key** for the hosted embedding and rerank services at `ai.mongodb.com`.
+- A **Claude API key**, or an Anthropic-compatible gateway via `ANTHROPIC_BASE_URL`.
+- **Node.js 20+** for the MCP set only. The MCP Server runs on Node.
+
+## Running locally
+
+```bash
+cd partners/langchain
+pip install -r requirements.txt
+cp .env.example .env      # then fill in your connection string and keys
+jupyter lab
+```
+
+`requirements.txt` pins every version, verified together on Python 3.12. The notebooks read
+credentials from `partners/langchain/.env` when there is one and prompt for anything missing,
+which is what lets the same notebook run in Colab or on your machine.
+
+Or click the **Open in Colab** badge at the top of any notebook and skip the install.
+
+## The index budget
+
+An M0 cluster allows three Atlas Search indexes, and these notebooks are built to fit.
+Notebook 1 of Advanced RAG creates `movies_vector_index`, notebook 2 creates
+`movies_fulltext_index`, and notebook 3 creates a short-lived index for the auto-embedding
+demo that it drops when it finishes. Each creation is guarded by an existence check, so
+re-running is free.
+
+At the limit they stop and list what is occupying the slots rather than freeing one: an
+index these notebooks did not create might be behind something you depend on, so asking
+for a fresh cluster is the safe failure. Notebook 3's scratch collection works the same
+way -- it refuses to reuse a collection it did not create, and
+`LANGCHAIN_AUTO_EMBED_COLLECTION` renames it if the default is taken on your cluster.
+
+The Mastra set uses the same two index names and also creates them if absent. The two sets
+are independent — run either first — and sharing the names is what keeps the pair inside the
+M0 budget. On a paid tier the limit is not binding.
+
+If you hit an index limit, look for leftovers from an interrupted run: the
+`auto_embed_demo` collection, or a demo index that outlived its notebook. Both are safe
+for you to drop. No notebook here deletes a collection or an index it did not create.