Skip to content

Home

Jump to bottom
Niccanor Dhas edited this page Feb 22, 2026 · 1 revision

Welcome to the tmam Wiki

tmam is an open-source, OpenTelemetry-native observability platform for AI applications. It automatically instruments LLM calls, agent workflows, vector database queries, and GPU metrics — and streams all that data into a powerful real-time dashboard.

You can run tmam fully on your own infrastructure with Docker, or use the managed cloud at cloud.tmam.ai.

📖 Wiki Pages

🚀 Getting Started

🔌 SDK Reference

🏢 Platform

  • Organizations & Teams — Create orgs, invite members, manage roles
  • Projects — Organize observability data within an organization
  • API Keys — Generate and manage SDK authentication keys

📊 Dashboard Features

🔐 Security & Auth

🤝 Contributing

Architecture Overview

┌─────────────────────────────────────────┐
│           Your Application              │
│                                         │
│  from tmam import init                  │
│  init(url=..., public_key=..., ...)     │
│                                         │
│  Auto-instrumented: OpenAI, LangChain,  │
│  CrewAI, Chroma, GPU metrics, etc.      │
└───────────────┬─────────────────────────┘
                │  OTLP (traces + metrics)
                ▼
┌───────────────────────────────────────────────────────┐
│                   tmam Server  :5050                  │
│                                                       │
│  /api/sdk  ← SDK endpoint (public/secret key auth)   │
│  /api/*    ← Dashboard API (JWT auth)                 │
│                                                       │
│  Parses protobuf traces, stores spans, metrics        │
└───────────────┬───────────────────────────────────────┘
                │
                ▼
┌──────────────────────┐     ┌─────────────────────────┐
│  MongoDB  :27018     │     │  Next.js Dashboard :3001 │
│                      │     │                         │
│  traces, metrics,    │     │  Observations, Analytics │
│  prompts, orgs,      │     │  Guardrails, Prompts,    │
│  guardrails, etc.    │     │  Evaluations, Settings   │
└──────────────────────┘     └─────────────────────────┘

Quick Links

Resource | URL -- | -- Cloud Platform | cloud.tmam.ai Documentation | docs.tmam.ai Python SDK (PyPI) | pypi.org/project/tmam GitHub Repository | github.com/tmam-dev/tmam

Welcome to the tmam Wiki

tmam is an open-source, OpenTelemetry-native observability platform for AI applications. It automatically instruments LLM calls, agent workflows, vector database queries, and GPU metrics — and streams all that data into a powerful real-time dashboard.

You can run tmam fully on your own infrastructure with Docker, or use the managed cloud at [cloud.tmam.ai](https://cloud.tmam.ai).

📖 Wiki Pages

🚀 Getting Started

🔌 SDK Reference

  • [SDK Initialization](SDK-Initialization) — init() options and configuration
  • [Supported Integrations](Supported-Integrations) — All 40+ auto-instrumented providers
  • [Manual Tracing](Manual-Tracing) — @trace decorator and start_trace() context manager
  • [GPU Monitoring](GPU-Monitoring) — Nvidia and AMD GPU metrics
  • [Prompt Hub](Prompt-Hub) — Fetch versioned prompts from the SDK
  • [Vault (Secrets)](Vault) — Encrypted secret storage and SDK retrieval
  • [Evaluations SDK](Evaluations-SDK) — Hallucination, toxicity, and bias detection
  • [Guardrails SDK](Guardrails-SDK) — Runtime input/output safety detection

🏢 Platform

  • [Organizations & Teams](Organizations-and-Teams) — Create orgs, invite members, manage roles
  • [Projects](Projects) — Organize observability data within an organization
  • [API Keys](API-Keys) — Generate and manage SDK authentication keys

📊 Dashboard Features

  • [Observations & Tracing](Observations-and-Tracing) — Traces, spans, requests, and exceptions
  • [Analytics](Analytics) — LLM, GPU, vector DB, scores, and model dashboards
  • [Prompt Management](Prompt-Management) — Version prompts and compare models in OpenGround
  • [Guardrails Dashboard](Guardrails-Dashboard) — Create and manage guardrail rules
  • [Evaluations Dashboard](Evaluations-Dashboard) — Datasets, experiments, and AI-arbiter scoring

🔐 Security & Auth

🤝 Contributing

  • [Contributing Guide](Contributing) — How to set up a dev environment and submit changes

Architecture Overview

┌─────────────────────────────────────────┐
│           Your Application              │
│                                         │
│  from tmam import init                  │
│  init(url=..., public_key=..., ...)     │
│                                         │
│  Auto-instrumented: OpenAI, LangChain,  │
│  CrewAI, Chroma, GPU metrics, etc.      │
└───────────────┬─────────────────────────┘
                │  OTLP (traces + metrics)
                ▼
┌───────────────────────────────────────────────────────┐
│                   tmam Server  :5050                  │
│                                                       │
│  /api/sdk  ← SDK endpoint (public/secret key auth)   │
│  /api/*    ← Dashboard API (JWT auth)                 │
│                                                       │
│  Parses protobuf traces, stores spans, metrics        │
└───────────────┬───────────────────────────────────────┘
                │
                ▼
┌──────────────────────┐     ┌─────────────────────────┐
│  MongoDB  :27018     │     │  Next.js Dashboard :3001 │
│                      │     │                         │
│  traces, metrics,    │     │  Observations, Analytics │
│  prompts, orgs,      │     │  Guardrails, Prompts,    │
│  guardrails, etc.    │     │  Evaluations, Settings   │
└──────────────────────┘     └─────────────────────────┘

Quick Links

Resource URL
Cloud Platform [cloud.tmam.ai](https://cloud.tmam.ai)
Documentation [docs.tmam.ai](https://docs.tmam.ai)
Python SDK (PyPI) [pypi.org/project/tmam](https://pypi.org/project/tmam)
GitHub Repository [github.com/tmam-dev/tmam](https://github.com/tmam-dev/tmam)

Clone this wiki locally