All Docs
FeaturesSidekickUpdated March 11, 2026

Semantic Memory & the OpenAI Embeddings Requirement

Semantic Memory & the OpenAI Embeddings Requirement

Sidekick's agent uses semantic memory search to recall relevant past context when handling tasks. This feature relies on vector similarity — comparing the numerical "meaning" of text snippets — to surface the most relevant memories from the agent's long-term store.

How It Works

  1. When the agent processes new information (messages, emails, events, etc.), it generates a vector embedding of that content using the OpenAI Embeddings API.
  2. These embeddings are stored in a vector index alongside the original content.
  3. At query time, the agent embeds the current context and performs a cosine similarity search to retrieve the most semantically relevant memories.
  4. Retrieved memories are injected into the agent's context window, giving it continuity across sessions and tasks.

Configuration

Semantic memory requires OPENAI_API_KEY to be set in the runtime environment:

OPENAI_API_KEY=sk-...

This key is used only for embedding generation (e.g. text-embedding-3-small or equivalent). It is not used for chat completion or other LLM inference.

Where to Get It

  1. Log in to platform.openai.com.
  2. Navigate to API Keys.
  3. Create a new secret key and copy it.
  4. Set it as OPENAI_API_KEY in your Sidekick environment configuration.

Graceful Degradation

If OPENAI_API_KEY is not set or is invalid:

  • Semantic memory search silently returns empty results — no error is thrown.
  • The agent continues to operate normally for all other tasks.
  • The agent simply has no access to long-term recall — it will not remember context from previous sessions.

This means the agent is still fully functional for stateless tasks (answering questions, taking single-turn actions, etc.) but will lose contextual continuity over time.

Recommended Setup

For any production deployment where agent quality matters, OPENAI_API_KEY should be treated as a required environment variable. While the absence of this key is not a hard blocker, it substantially reduces the agent's ability to:

  • Remember user preferences and past instructions
  • Recognise recurring tasks and patterns
  • Maintain continuity across long-running workflows
  • Provide personalised, context-aware responses

Summary

StateSemantic MemoryAgent Function
OPENAI_API_KEY set✅ Fully operational✅ Full quality
OPENAI_API_KEY absent❌ Empty results (silent)⚠️ Degraded recall