12 files changed
@@ -55,6 +55,7 @@ | |||
| 55 | 55 | * [Retrieval Augmented Generation (RAG) with Feast](tutorials/rag-with-docling.md) | |
| 56 | 56 | * [RAG Fine Tuning with Feast and Milvus](../examples/rag-retriever/README.md) | |
| 57 | 57 | * [MCP - AI Agent Example](../examples/mcp_feature_store/README.md) | |
| 58 | + * [Feast-Powered AI Agent](../examples/agent_feature_store/README.md) | ||
| 58 | 59 | ||
| 59 | 60 | ## How-to Guides | |
| 60 | 61 | ||
@@ -89,6 +89,17 @@ Implement semantic search by: | |||
| 89 | 89 | 2. Converting search queries to embeddings | |
| 90 | 90 | 3. Finding semantically similar documents using vector search | |
| 91 | 91 | ||
| 92 | + ### AI Agents with Context and Memory | ||
| 93 | + | ||
| 94 | + Feast can serve as both the **context provider** and **persistent memory layer** for AI agents. Unlike stateless RAG pipelines, agents make autonomous decisions about which tools to call and can write state back to the feature store: | ||
| 95 | + | ||
| 96 | + 1. **Structured context**: Retrieve customer profiles, account data, and other entity-keyed features | ||
| 97 | + 2. **Knowledge retrieval**: Search vector embeddings for relevant documents | ||
| 98 | + 3. **Persistent memory**: Store and recall per-entity interaction history (last topic, resolution, preferences) using `write_to_online_store` | ||
| 99 | + 4. **Governed access**: All reads and writes are subject to the same RBAC, TTL, and audit policies as any other feature | ||
| 100 | + | ||
| 101 | + With MCP enabled, agents built with any framework (LangChain, LlamaIndex, CrewAI, AutoGen, or custom) can discover and call Feast tools dynamically. See the [Feast-Powered AI Agent example](../../examples/agent_feature_store/) and the blog post [Building AI Agents with Feast](https://feast.dev/blog/feast-agents-mcp/) for a complete walkthrough. | ||
| 102 | + | ||
| 92 | 103 | ### Scaling with Spark Integration | |
| 93 | 104 | ||
| 94 | 105 | Feast integrates with Apache Spark to enable large-scale processing of unstructured data for GenAI applications: | |
@@ -167,9 +178,11 @@ The MCP integration uses the `fastapi_mcp` library to automatically transform yo | |||
| 167 | 178 | The fastapi_mcp integration automatically exposes your Feast feature server's FastAPI endpoints as MCP tools. This means AI assistants can: | |
| 168 | 179 | ||
| 169 | 180 | * **Call `/get-online-features`** to retrieve features from the feature store | |
| 181 | + * **Call `/retrieve-online-documents`** to perform vector similarity search | ||
| 182 | + * **Call `/write-to-online-store`** to persist agent state (memory, notes, interaction history) | ||
| 170 | 183 | * **Use `/health`** to check server status | |
| 171 | 184 | ||
| 172 | - For a complete example, see the [MCP Feature Store Example](../../examples/mcp_feature_store/). | ||
| 185 | + For a basic MCP example, see the [MCP Feature Store Example](../../examples/mcp_feature_store/). For a full agent with persistent memory, see the [Feast-Powered AI Agent Example](../../examples/agent_feature_store/). | ||
| 173 | 186 | ||
| 174 | 187 | ## Learn More | |
| 175 | 188 | ||
@@ -181,6 +194,8 @@ For more detailed information and examples: | |||
| 181 | 194 | * [Milvus Quickstart Example](https://github.com/feast-dev/feast/tree/master/examples/rag/milvus-quickstart.ipynb) | |
| 182 | 195 | * [Feast + Ray: Distributed Processing for RAG Applications](https://feast.dev/blog/feast-ray-distributed-processing/) | |
| 183 | 196 | * [MCP Feature Store Example](../../examples/mcp_feature_store/) | |
| 197 | + * [Feast-Powered AI Agent Example (with Memory)](../../examples/agent_feature_store/) | ||
| 198 | + * [Blog: Building AI Agents with Feast](https://feast.dev/blog/feast-agents-mcp/) | ||
| 184 | 199 | * [MCP Feature Server Reference](../reference/feature-servers/mcp-feature-server.md) | |
| 185 | 200 | * [Spark Data Source](../reference/data-sources/spark.md) | |
| 186 | 201 | * [Spark Offline Store](../reference/offline-stores/spark.md) | |
0 commit comments