Tutorials & How‑Tos Kay x Cybersyn x LangChain: Embedding SEC Filings for RAG
Integrate KayAiRetriever into your LangChain app to fetch SEC filings via hosted embeddings, specifying dataset_id='company', data_types=['10-K','10-Q'], num_contexts=6.
Install Kay via pip, obtain API key, and instantiate KayAiRetriever with dataset_id='company', data_types=['10-K','10-Q'], num_contexts=6 to start retrieving SEC filing context in your LangChain agent.
Summary
Kay and Cybersyn partnered with LangChain to launch the SEC Retriever on LangChain, a retrieval‑augmented generation (RAG) solution that pulls context from U.S. public company filings. The service, available as a pip package (pip install kay), requires a free API key from Kay.ai and uses the KayAiRetriever class to index datasets such as 10‑K and 10‑Q documents. The retriever returns the top six chunks for a query, and the data is sourced from Snowflake Marketplace via Cybersyn’s LLM‑training‑essentials product, which converts raw SEC filings into relational tables with company metadata. The article demonstrates a sample query that extracts financial patterns for Nvidia over three quarters, highlighting the precision of pre‑embedded, enriched data. The solution eliminates the need for custom OCR or embedding pipelines, and supports real‑time RAG for conversational agents.
The SEC Retriever is positioned as a cost‑effective, scalable alternative to training on derivative articles, and it is designed to handle the complex formats of SEC filings, including tables and XBRL data. By providing fully hosted embeddings and a semantic retrieval API, Kay and Cybersyn enable developers to focus on prompt design rather than data ingestion.
Key takeaways include the partnership model, the use of Snowflake for data governance, and the ability to retrieve context with minimal setup.
Key changes
- Kay and Cybersyn launch SEC Retriever on LangChain
- Provides fully hosted embeddings for SEC filings
- Supports dataset_id 'company' with data_types 10‑K and 10‑Q
- Returns top 6 chunks per query
- Uses Kay API key via pip install kay
- Data sourced from Snowflake Marketplace
- Example query returns financial patterns for Nvidia
- Embeddings enriched with company metadata