I Gave Two TypeScript Agent Sessions One Shared Memory With Hindsight (No LLM Key)

Description and use

Hindsight enables agents to persist and retrieve repository knowledge without an LLM key, speeding up subsequent sessions.

Problem

Agents lose contextual knowledge between sessions, leading to repeated discovery of repo conventions and hidden constraints.

Outcome

Hindsight successfully stored and retrieved five facts about the shop‑api repo in no‑LLM mode, with zero token usage and fast query times.

Workflow

• Install Hindsight API server via pip in a Python 3.12 virtual environment
• Configure Hindsight to run in no‑LLM mode with HINDSIGHT_API_LLM_PROVIDER=none
• Start the server and note startup times and resource usage
• Create a TypeScript client with @vectorize-io/hindsight-client
• In Agent A script, create a bank and retain five domain facts
• In Agent B script, instantiate a new client and query the bank with four questions
• Observe that answers are retrieved from stored facts without LLM usage

Evidence

Hindsight documents a mode for exactly that, so this is a test of its no-LLM floor, not of everything it can do.

Source

The five retains took 184, 42, 46, 56 and 41 ms. Every response reported zero LLM tokens, which is expected with no model.

Source

The startup banner showed v0.10.2 with an embedded PostgreSQL (pg0, started automatically), LLM none / none, local embeddings (BAAI/bge-small-en-v1.5), a local reranker (cross-encoder/ms-marco-MiniLM-L-6-v2), and MCP enabled at /mcp.

Source