API Documentation

The LongMemory API allows you to add infinite semantic memory to your AI agents. We provide a unified layer that handles vector embeddings, storage, and retrieval automatically.

Base URL

https://api.longmemory.io/v4

Version

v1.0

Authentication

Authentication is handled via Bearer tokens. You can obtain your API key from the dashboard. Always keep your keys secure and never expose them in client-side code.

Python SDK

The client automatically looks for these environment variables.

bash

# In your terminal or .env file
export LONGMEMORY_API_KEY="sk_live_12345..."

POST

/ingest

Stores a new memory chunk. The system automatically processes the text, generates vector embeddings using our SOTA models, and indexes it for low latency retrieval.

Parameters

Property	Type	Description
user_id Required	string	Unique identifier for the user or agent owner.
text Required	string	The conversation content or fact to be stored.
speaker Optional	string	Name of the speaker (e.g., 'Bob', 'Alice').
created_at Optional	iso8601	Timestamp for historical backfilling (e.g., '2023-10-01T12:00:00Z').

Python Client

python

client.add(
    user_id="user_123",
    text="My favorite food is Pizza.",
    speaker="Bob"
)

cURL Request

bash

curl -X POST https://api.longmemory.io/v4/ingest \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "user_123",
    "text": "My favorite food is Pizza.",
    "speaker": "Bob"
  }'

Response

json

{
  "id": "user_123...",
  "status": "success",
  "token_usage": 12
}

POST

/query

Parameters

Property	Type	Description
user_id Required	string	Unique identifier for the user.
query Required	string	The question or statement to retrieve context for.

Python Client

python

# Retrieve the memories for Bob
results = client.get(
    user_id="user_123",
    query="What is Bob's favorite food?",

)

cURL Request

bash

curl -X POST https://api.longmemory.io/v4/query \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "user_123",
    "query": "What is Bob's favorite food?",

  }'

Response

json

{
  "matches": [
    {
      "text": "[Bob] My favorite food is Pizza.",
    }
  ]
}

POST

/askCOMING SOON

Let LongMemory handle the LLM generation for you. We securely proxy the request, retrieve the relevant context, and prompt the LLM of your choice in one step.

Parameters

Property	Type	Description
user_id Required	string	Unique identifier for the user.
query Required	string	The question to ask the agent.
llm_model Required	string	The model ID (e.g., 'gpt-4o', 'claude-3').
llm_api_key Required	string	Your LLM provider's API key (we do not store this).

Python Client

python

# 1. Initialize with your LLM keys
client = LongMemory(
    api_key="lm_live_...",
    llm_api_key="sk-openai-...",
    llm_model="gpt-4o"
)

# 2. Ask the agent directly
response = client.ask(
    user_id="user_123",
    query="What is the name of Alice's pet?"
)

cURL Request

bash

curl -X POST https://api.longmemory.io/v4/ask \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "user_123",
    "query": "What is the name of Alice's pet?",
    "llm_model": "gpt-4o",
    "llm_api_key": "sk-openai-..."
  }'

Response

json

{
  "answer": "Alice's pet is named Fluffy.",
  "context_used": [
    "[Alice] I have a cat named Fluffy."
  ]
}