abidex

Abidex

Zero-code monitoring and tracing for AI agents. Add one import, get spans for workflows, agents, and tasks—with role, goal, backstory, and timing out of the box.

---

Installation

pip install abidex

For sending traces to an OTLP backend (SigNoz, Uptrace, Jaeger, etc.):

pip install abidex[otlp]

Optional dev setup (editable install + lint/test):

pip install -e ".[dev]"

Also works with uv, pdm, hatch: uv add abidex, pdm add abidex, etc.

---

Quickstart

1. Install: pip install abidex
2. Import abidex before your crew/graph so it can patch frameworks.
3. Run your agent script as usual—spans are created automatically.

import abidex  # ← add this first

from crewai import Agent, Task, Crew

researcher = Agent(role="Researcher", goal="Find facts", backstory="You are a researcher.")
task = Task(description="Summarize the topic", agent=researcher)
crew = Crew(agents=[researcher], tasks=[task])

result = crew.kickoff(inputs={"topic": "OpenTelemetry"})

No config required. Traces go to the console by default; point OTEL_EXPORTER_OTLP_ENDPOINT at a backend for a full UI.

---

Add Abidex to your agent script

Import abidex before you import or use CrewAI, LangGraph, or Pydantic AI. One line is enough.

CrewAI

import abidex

from crewai import Agent, Task, Crew

agent = Agent(role="Analyst", goal="Analyze data", backstory="You are an analyst.")
task = Task(description="Summarize the report", agent=agent)
crew = Crew(agents=[agent], tasks=[task])
result = crew.kickoff(inputs={"topic": "Q4 report"})

LangGraph

import abidex

from langgraph.graph import StateGraph, MessagesState
# ... build your graph ...
compiled = graph.compile()
result = compiled.invoke({"messages": [...]})

Pydantic AI

import abidex

from pydantic_ai import Agent

agent = Agent("my-model", system_prompt="You are a helpful assistant.")
result = agent.run_sync("Explain observability in one sentence.")

---

How to see traces

Console (default)

If you don’t set OTEL_EXPORTER_OTLP_ENDPOINT, spans are printed to stderr. Ideal for local debugging.

SigNoz (persistent UI + dashboards)

1. Start SigNoz (see examples/signoz-quickstart.md or SigNoz Docker docs):

   git clone https://github.com/signoz/signoz.git && cd signoz
   docker compose -f deploy/docker-compose/standalone/docker-compose.yaml up -d
   

2. Set env and run your script:

   export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
   pip install abidex[otlp]
   python your_agent_script.py
   

3. Open http://localhost:3301 → Traces. Filter by gen_ai.agent.role, gen_ai.workflow.name, etc.

Uptrace (lightweight, dev-friendly)

1. Start Uptrace:

   docker run -d -p 14317:4317 -p 14318:4318 --name uptrace \
     -e UPTRACE_DSN=postgres://uptrace:uptrace@host.docker.internal:5432/uptrace \
     uptrace/uptrace:latest
   

2. Set env and run:

   export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:14317
   pip install abidex[otlp]
   python your_agent_script.py
   

3. Open http://localhost:14318 → sign up (first user) → Traces.

See examples/uptrace-quickstart.md for more detail.

---

Testing checklist

After running a traced workflow, verify:

Example for tests:

ABIDEX_AUTO=false pytest tests/

---

CLI

The abidex command gives you a rich CLI (tables, colors, spinners). Enable the in-memory buffer so the CLI can see spans from the same process:

export ABIDEX_BUFFER_ENABLED=true
python your_agent_script.py
abidex trace last 10

Examples:

abidex status
abidex status -v
abidex trace last 10
abidex trace last 5 --filter "role=Researcher"
abidex trace last --file spans.ndjson
abidex trace export --format jsonl -o traces.ndjson --last 100
abidex trace export --format pretty --last 5
abidex init
abidex summary

For cross-process use: export from your app with abidex.trace_buffer.export_to_jsonl("spans.ndjson", 100), then run abidex trace last --file spans.ndjson.

---

Supported frameworks

---

Configuration

---

Troubleshooting

---

Contributing & feedback

• Bugs & features: GitHub Issues.
• Contributions: PRs welcome. See CONTRIBUTING.md for dev setup, tests, and how to add framework patches.

We’re focused on execution observability (workflow/agent/task spans and GenAI attributes). More frameworks and deeper instrumentation are on the roadmap.

License

This project is licensed under the Business Source License 1.1 (BSL 1.1). You may copy, modify, and redistribute the software for non-production use; production use requires a commercial license from Abide or compliance when the Change Date is reached. On the Change Date (see LICENSE), the license converts to the Apache License, Version 2.0. See LICENSE and MariaDB BSL 1.1 for full terms.