Mem0 Custom LLM Integration with Fikra API

Prerequisites

• A Fikra API key (fk_live_...) from the Fikra dashboard.
• Python 3.9+ or Node.js 18+.
• The mem0ai package installed (pip install mem0ai or npm install mem0ai).
• Access to the Fikra API base URL: https://api.fikraapi.co.ke/v1.

Environment Variable Setup

Set the following environment variables in your shell or .env file:

export FIKRA_API_KEY="fk_live_your_key_here"
export FIKRA_BASE_URL="https://api.fikraapi.co.ke/v1"

Python Example

Mem0 supports custom LLM providers via its LLM configuration. Since Fikra API is OpenAI-compatible, you can use the openai provider type with a custom base URL.

import os
from mem0 import Memory

# Configure Mem0 to use Fikra API via OpenAI-compatible interface
config = {
    "llm": {
        "provider": "openai",
        "config": {
            "model": "fikra-flash",
            "api_key": os.getenv("FIKRA_API_KEY"),
            "base_url": os.getenv("FIKRA_BASE_URL", "https://api.fikraapi.co.ke/v1")
        }
    },
    "embedder": {
        "provider": "openai",
        "config": {
            "model": "text-embedding-3-small",
            "api_key": os.getenv("FIKRA_API_KEY"),
            "base_url": os.getenv("FIKRA_BASE_URL", "https://api.fikraapi.co.ke/v1")
        }
    }
}

memory = Memory(config=config)

# Add a memory
memory.add("User prefers dark mode in the app", user_id="user_123")

# Search memories
results = memory.search("What are the user's UI preferences?", user_id="user_123")
print(results)

Note: The embedder configuration above assumes Fikra API supports OpenAI-compatible embedding endpoints. If Fikra API does not expose an embeddings endpoint, you must use a separate embedding provider (e.g., OpenAI, Cohere) for the embedder section while keeping the llm section pointed at Fikra. Check the Fikra model registry for available embedding models.

Streaming Guidance

Fikra API supports streaming via stream: true and Server-Sent Events (SSE). Mem0’s default LLM calls are non-streaming. If you need streaming for direct chat completions (not through Mem0’s memory operations), use the OpenAI SDK directly:

from openai import OpenAI
import os

client = OpenAI(
    api_key=os.getenv("FIKRA_API_KEY"),
    base_url=os.getenv("FIKRA_BASE_URL", "https://api.fikraapi.co.ke/v1")
)

stream = client.chat.completions.create(
    model="fikra-flash",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Common Errors

Production Checklist

• [ ] Store FIKRA_API_KEY in a secrets manager, not in source code.
• [ ] Set FIKRA_BASE_URL explicitly to avoid defaulting to OpenAI’s endpoint.
• [ ] Verify that the model alias (fikra-flash) is available in the Fikra model registry.
• [ ] Implement retry logic with exponential backoff for 429 and 503 responses.
• [ ] Monitor token-credit balance to avoid 402 errors in production.
• [ ] If using embeddings, confirm whether Fikra API provides an embeddings endpoint; otherwise, configure a separate embedding provider.
• [ ] Test streaming separately from Mem0’s memory operations if SSE support is required for user-facing chat.
• [ ] Log request IDs from Fikra API responses for support and debugging.