Feenion AI Debugger

Description

Feenion is an open-source AI debugger and observability platform designed to provide developers with comprehensive insights into AI systems. It offers full-DAG execution traces, token costs, latency flamegraphs, and error diagnostics specifically for LLMs, RAG pipelines, and autonomous multi-agent tool loops. Unlike traditional APMs, which only capture raw HTTP status codes or generic server errors, Feenion captures the full causal execution graph, including exact prompts, completions, tokens, calculated costs, tool arguments, and retrieval scores.

Feenion is built to address the challenges of debugging AI systems, which are often complex distributed systems involving prompts, vector databases, and multi-step tool reasoning. It provides developers with the ability to inspect inputs, outputs, tokens, latency, and costs for each component, offering a complete X-ray vision into every prompt, retrieval chunk, tool call, and token dollar. This is achieved through a lightweight container and a Python library that runs locally, ensuring data privacy and eliminating the need for cloud vendor lock-in.

The platform is equipped with features such as flamegraph timelines to identify slow vector lookups and model latencies, a D3 execution DAG to visualize complex causality graphs, and real-time financials to calculate exact token expenditures. Feenion also supports multi-currency token and cost tracking, allowing users to switch between different currencies and customize exchange rates.

Feenion is licensed under Apache 2.0, allowing users to inspect, modify, and extend any part of the codebase. It is designed to run on your own hardware without per-seat pricing or cloud vendor lock-in, making it an ideal choice for developers seeking a self-hosted, open-source alternative for AI observability.

Feenion AI Debugger's Core Features

• Full-DAG execution traces

• Token cost tracking

• Latency flamegraphs

• Error diagnostics

• Runs locally

• Open-source under Apache 2.0

• Multi-currency support

• D3 execution DAG visualization

• Real-time financials

• Flamegraph timelines

• Zero cloud vendor lock-in

• Python SDK integration

• Multi-tenant workspace isolation

• Semantic error intelligence

• Comparative trace regression diff

How to use Feenion AI Debugger?

1. 1

   Configure: Install the Python SDK and set up the client

2. 2

   Use: Run Feenion locally with Docker

3. 3

   Inspect: View execution traces and diagnostics

4. 4

   Optimize: Analyze latency and cost metrics

Feenion AI Debugger's Use Cases

• AI Debugging
• Cost Management
• Latency Analysis
• Data Privacy
• Open-Source Development

FAQ from Feenion AI Debugger

• What is Feenion AI Debugger?

  Feenion is an open-source AI debugger and observability platform for LLMs, RAG pipelines, and multi-agent tool loops.

• How does Feenion ensure data privacy?

  Feenion runs locally on your machine, ensuring that all telemetry data stays under your control without cloud vendor lock-in.

• What are the key features of Feenion?

  Key features include full-DAG execution traces, token cost tracking, latency flamegraphs, error diagnostics, and real-time financials.

• Is Feenion open-source?

  Yes, Feenion is open-source and licensed under Apache 2.0, allowing users to inspect, modify, and extend the codebase.

• Can Feenion be used for production-scale AI applications?

  Yes, Feenion is designed for production scale, providing tools to inspect, benchmark, and debug AI workloads.

• Does Feenion support multi-currency token tracking?

  Yes, Feenion supports multi-currency token and cost tracking with customizable exchange rates.

• What is the deployment model for Feenion?

  Feenion can be deployed as a single lightweight container using Docker, with support for both SQLite and PostgreSQL.

• Who developed Feenion?

  Feenion was developed by Darshan as an engineering project to explore distributed tracing in AI applications.