Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.thrindex.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Memory

A memory is a text fact tied to an agent and a user. You send content (max 8 KB), plus agent_id and user_id. Optional: metadata (JSON), confidence (0–1, default 1). Write short, factual sentences — one idea per memory when you can. 
User prefers weekly email digests.
User timezone is Europe/Berlin.

​
Scoping

ScopeSource
OrganizationFrom your API key — never sent in the request body
agent_idYou — identifies which agent owns the memory
user_idYou — your app’s user id
Search only returns memories for the agent_id and user_id you pass. Keys cannot access another organization’s data. agent_id and user_id: required on write and search, max 256 chars, charset [A-Za-z0-9._:-].

​
Write

POST /v1/memories → 202 with { "id": "...", "status": "queued" }. Processing (embedding, ranking metadata) runs in the background. Do not assume the memory is searchable in the same millisecond. POST /v1/memories/search → 200 with ranked results.
FieldDefaultNotes
queryrequiredNatural language, max 2048 chars
k101–100
task_context—Optional; improves ranking for the current task
levelall0, 1, or 2 to filter
Each result includes content, score, importance, and created_at.

​
Typical loop

  1. Search for context when a session starts or before an LLM call.
  2. After a turn worth remembering, call add.
  3. Inject search results into the prompt.
Call the API from your backend only. Do not put API keys in browsers or mobile clients.

​
Delete

DELETE /v1/memories/{id} — soft delete by default (excluded from search). DELETE /v1/memories/{id}?hard=true — permanent erasure (GDPR). Irreversible.