Quickstart

1. Create a project & API key

From the dashboard, create an organization, then a project, then an API key. The key is shown once — store it as THETA_API_KEY. Project IDs are optional in SDK code because API keys are scoped to a project server-side. Set THETA_PROJECT only when you want explicit project filtering.

2. Install an SDK

pip install theta-obsrv

3. Send a trace

import os
from theta_observability import TraceClient

client = TraceClient(
    api_key=os.environ["THETA_API_KEY"],
)

with client.trace(name="hello-world", run_type="dev") as t:
    with t.step(name="greet", type="llm", model="claude-sonnet-4-6") as s:
        s.log_message(role="user", text="Say hi to obsrv")
        s.log_message(role="assistant", text="Hi, obsrv!")
        s.set_token_usage(input=12, output=4)

client.flush()

4. See it in the dashboard

Open app.obsrv.tech. The trace should be visible in the live tail within seconds. Click in for the full step timeline, messages, and metadata.

5. Where to go next

• Wrap your real agent with client.wrap_agent() to capture every run automatically.
• Add an OpenAI or Anthropic adapter so model calls appear as steps without you logging them.
• Attach images, audio, video, sensor frames, or screenshots to steps. See Attachments.
• Tag traces by release with set_metadata(release="v3.2.1") and use that to compare regressions.