Quick Start
Fastmem — Memory and Context Engineering for AI Apps
Give your AI app long-term memory and task-ready context. Fastmem extracts facts from conversations, tracks structured state and constraints, and can build the final context packet your app sends to an LLM.
pip install fastmemfrom fastmem import FastmemClient
memory = FastmemClient(
api_key="your-api-key"
)# Add memories - facts are automatically extracted
result = memory.add(
user_id="user_123",
text="I'm allergic to peanuts and I work at Google"
)
print(f"Added {result.memory_count} memories")# Search for relevant memories
results = memory.search(
user_id="user_123",
query="What food allergies does this user have?"
)
for mem in results:
print(f"- {mem.text} (score: {mem.score:.2f})")import requests
resp = requests.post(
"https://api.fastmem.ai/api/context/build",
headers={"X-API-Key": "your-api-key", "Content-Type": "application/json"},
json={
"user_id": "user_123",
"task_id": "camera_shopping",
"query": "What is the price of Canon R6 and is it within my budget?",
"context": "shopping",
"output_format": "markdown"
}
)
packet = resp.json()
print(packet["formatted_context"])
print(packet["summary"])Use the primitive APIs when you want raw memories. Use the context builder when you want Fastmem to assemble the structured packet for the next LLM call.
AI Chatbots
Your chatbot remembers user preferences, past conversations, and personal details across sessions.
Context buckets keep work/personal separate
Shopping Agents
Track budgets, constraints, and preferences. Build context packets with exact prices, rules, and semantic facts.
Structured state + constraints + verbatim values
Health & CRM Apps
Multi-party memory with speaker roles. Track what patients/clients say vs. what professionals note.
Tenant isolation + speaker roles + audit trail
Fastmem sits between your app and the LLM. It stores memories, your app generates responses.
Ready to try it?
Test memory extraction, search, and context building in the Playground
Returns the structured packet and formatted context your app would pass into an LLM.
Select or enter a user to see their memories
| Prefix | Memories | Users |
|---|
| User ID | Memories |
|---|
Loading users...
Select a user to view memories
| Context | Memories | Users | Latest Activity |
|---|
| # | Memory | Category | Cosine | Decay | Final Score | Info |
|---|
Enter any text to see how Fastmem classifies it — temporal type, TTL, and salience.
Evolution Timeline
Track how user preferences and facts change over time
Select a user and click "Load Insights" to see patterns
API Reference
Complete reference for the Fastmem REST API
/api/add
Add memories from text. Automatically extracts facts using AI.
Request Body
{
"user_id": "string", // Required: Unique user identifier
"text": "string", // Required: Text to extract memories from
"context": "string" // Optional: Context bucket (default: "default")
}import requests
response = requests.post(
"https://api.fastmem.ai/api/add",
headers={"X-API-Key": "your-api-key"},
json={
"user_id": "user_123",
"text": "I'm allergic to peanuts",
"context": "health"
}
)
print(response.json())/api/search
Search for relevant memories using semantic similarity.
Query Parameters
user_id (required) - User identifierquery (required) - Search querytop_k (optional) - Max results (default: 5)context (optional) - Filter by contextresponse = requests.get(
"https://api.fastmem.ai/api/search",
headers={"X-API-Key": "your-api-key"},
params={
"user_id": "user_123",
"query": "What allergies?",
"top_k": 5
}
)
for result in response.json()["results"]:
print(f"{result['text']} - {result['score']:.2f}")/api/memories
List all memories for a user with optional filtering.
Query Parameters
user_id (required) - User identifiercontext (optional) - Filter by contextcategory (optional) - Filter by categoryView all endpoints in the interactive API documentation
Open Swagger UINo logs yet
Make some API calls to see logs here
Provide your own OpenAI API key for unlimited usage. Without it, you're limited to trial credits.
Loading keys...