phylax 1.0.5

Deterministic regression enforcement for LLM systems.

Phylax

Deterministic regression enforcement for LLM systems.

---

The Problem

LLM outputs change unexpectedly. Same prompt, different model version → different behavior. Without Phylax, you discover this in production.

Installation

pip install phylax

For server/UI support:

pip install phylax[server]

For all LLM providers:

pip install phylax[all]

Quick Start

import phylax
from phylax._internal.decorator import trace
from phylax._internal.context import execution
from phylax._internal.adapters.gemini import GeminiAdapter

# add google api key
@trace(provider="gemini")
def ask_gemini(prompt: str):
    """Traced Gemini call."""
    adapter = GeminiAdapter()
    response, _ = adapter.generate(
        prompt=prompt,
        model="gemini-2.5-flash",
    )
    return response

# Single call
result = ask_gemini("Hello!")
print(result.text)

# Track multi-step agent flows
with execution() as exec_id:
    step1 = ask_gemini("What is 2+2?")
    step2 = ask_gemini("Is that correct?")

# Start the UI server
phylax server
# Open http://127.0.0.1:8000/ui

# Mark a known-good response as baseline
phylax bless <trace_id>

# In CI: fail if output regresses
phylax check  # exits 1 on failure

That's it. Your CI now blocks LLM regressions.

---

What Phylax is NOT

• ❌ Not monitoring — no metrics, no dashboards
• ❌ Not observability — no traces-to-cloud, no analytics
• ❌ Not AI judgment — rules are deterministic, not LLM-based
• ❌ Not cloud-dependent — runs entirely local
• ❌ Not prompt engineering — tests outputs, not prompts

Phylax is a test framework. It tells you when LLM behavior changes.

---

Using the Web UI

pip install phylax[server]
phylax server
# Open http://127.0.0.1:8000/ui

The UI shows:

• Trace list with pass/fail status
• Execution graphs for multi-step flows
• Forensics mode for debugging

---

CI Integration

# .github/workflows/phylax.yml
- run: phylax check
  env:
    GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}

Exit codes:

• 0 — All golden traces pass
• 1 — Regression detected

---

Commands

Command	What it does
phylax init	Initialize config
phylax server	Start API server + UI
phylax list	List traces
phylax list --failed	Show only failed traces
phylax show <id>	Show trace details
phylax replay <id>	Re-run a trace
phylax bless <id>	Mark as golden baseline
phylax check	CI regression check

---

Features

Feature	Description
Trace Capture	Record every LLM call automatically
Execution Context	Group traces by execution() context
Golden Traces	Baseline comparisons with hash verification
CI Integration	phylax check exits 1 on regression
Web UI	View traces at http://127.0.0.1:8000/ui
Forensics Mode	Debug failures with guided investigation

---

Stability Guarantee

Phylax v1.x is API-frozen:

• No breaking changes in v1.x
• trace, execution are stable
• Exit codes are stable
• Schema is stable

See docs/contract.md for full guarantees.

---

Documentation

• Quickstart
• Mental Model
• API Contract

---

License

MIT License