tracerazor-openai-agents
OpenAI Agents SDK hooks adapter for TraceRazor.
Automatically captures every LLM call, tool execution, and agent handoff
from a Runner.run() invocation — zero manual instrumentation required.
v0.2.0 — New Metrics:
- ✨ Semantic Continuity (CSD) — Detects when your agent's reasoning drifts topic mid-execution
- ✨ Adherence Scoring (IAR) — After optimizing, validates that fixes actually improved metrics
Install
pip install tracerazor-openai-agents
pip install tracerazor-openai-agents[agents] # includes openai-agents
Requires the tracerazor binary:
cargo build --release
export TRACERAZOR_BIN=/path/to/TraceRazor/target/release/tracerazor
Usage
from tracerazor_openai_agents import TraceRazorHooks
from agents import Agent, Runner
hooks = TraceRazorHooks(
agent_name="support-agent",
threshold=70,
)
result = await Runner.run(
agent,
"I need a refund for order ORD-9182",
hooks=hooks,
)
# After the run:
report = hooks.analyse()
print(report.markdown())
# CI/CD gate — raises AssertionError if TAS < threshold:
hooks.assert_passes()
API
TraceRazorHooks(agent_name, framework, threshold, task_value_score, tracerazor_bin)
| param | default | description |
|---|---|---|
agent_name |
"openai-agent" |
shown in all reports |
framework |
"openai-agents" |
framework label |
threshold |
70.0 |
minimum TAS for assert_passes() |
task_value_score |
1.0 |
answer quality (0–1) |
tracerazor_bin |
auto | path to binary; falls back to TRACERAZOR_BIN env var |
hooks.analyse() → TraceRazorReport
Finalise and submit the trace. Returns the report.
hooks.assert_passes()
Raise AssertionError if TAS < threshold.
hooks.set_task_value_score(score: float)
Update quality score before calling analyse().
Captured events
| SDK hook | TraceRazor step type |
|---|---|
on_agent_end |
reasoning (agent output) |
on_tool_start / on_tool_end |
tool_call (success) |
on_handoff |
reasoning (handoff marker) |
Multi-agent traces
When your workflow uses multiple agents via handoffs, TraceRazor automatically produces a per-agent breakdown alongside the composite score:
MULTI-AGENT BREAKDOWN
Agent Steps Tokens Share TAS Grade
TriageAgent 4 1,200 28.6% 82.5 GOOD
SupportAgent 7 2,600 61.9% 61.2 FAIR
EscalationAgent 2 400 9.5% N/A N/A
The composite TAS is weighted by each agent's token consumption.
See Multi-Agent Guide for detailed workflow examples with cost analysis and optimization validation.
Metadata
Release files for tracerazor-openai-agents 1.0.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| tracerazor_openai_agents-1.0.0.tar.gz | 8.9 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| tracerazor_openai_agents-1.0.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 17.4 kB
Release files / tracerazor_openai_agents-1.0.0.tar.gz
| Download URL | tracerazor_openai_agents-1.0.0.tar.gz |
|---|---|
| Size | 8.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
baec210434d13c7e63a0febd8ed12508420d672b54012fd365d2f88fb9e229d8
|
|
BLAKE2b-256 checksum How to use checksums |
16f1c7c52845d86715c36676eeae997b3ead573ace27555d8951a544149877af
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.11.6
|
Release files / tracerazor_openai_agents-1.0.0-py3-none-any.whl
| Download URL | tracerazor_openai_agents-1.0.0-py3-none-any.whl |
|---|---|
| Size | 8.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
9268efd258d91716026554f546c2354a350a8c12a3f07678a4ee9c2c606ea3e7
|
|
BLAKE2b-256 checksum How to use checksums |
aef7cc58bfd394890bbe0974c5b303964519374e29d3470f6a7a0e35a02633dd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.11.6
|