Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion module_2/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ Full setup with Slack and Gmail (reads credentials from your shell environment):
```bash
docker run -it --rm \
-p 8501:8501 -p 8502:8502 \
-e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
-e OPENROUTER_API_KEY=$OPENROUTER_API_KEY \
-e SLACK_BOT_TOKEN=$SLACK_BOT_TOKEN \
-e SLACK_TEAM_ID=$SLACK_TEAM_ID \
-v "$PWD":/workspace \
Expand Down
2 changes: 1 addition & 1 deletion module_3/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -46,7 +46,7 @@ docker run -it --rm \
-p 8501:8501 -p 8502:8502 \
-p 6274:6274 -p 3000:3000 \
-p 8001:8001 -p 8002:8002 \
-e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
-e OPENROUTER_API_KEY=$OPENROUTER_API_KEY \
-e SLACK_BOT_TOKEN=$SLACK_BOT_TOKEN \
-e SLACK_TEAM_ID=$SLACK_TEAM_ID \
-v "$PWD":/workspace \
Expand Down
12 changes: 8 additions & 4 deletions module_3/eval/orchestrator.py
Original file line number Diff line number Diff line change
Expand Up @@ -2,8 +2,9 @@
"""
Real LLM orchestrator for Module 3 eval harness.
Runs a multi-agent pipeline with actual Claude API calls and writes the
transcript JSON + audit log that test_deterministic.py expects.
Runs a multi-agent pipeline with actual LLM calls (Claude models served via
OpenRouter's Anthropic-compatible endpoint) and writes the transcript JSON +
audit log that test_deterministic.py expects.
Usage (from module_3/):
python3 eval/orchestrator.py [options]
Expand Down Expand Up @@ -46,7 +47,7 @@

# ── model ────────────────────────────────────────────────────────────────────

MODEL = "claude-sonnet-4-6"
MODEL = "anthropic/claude-sonnet-4.6"

# ── tool grant map (mirrors docs/routing-and-tool-grant-map.json) ─────────────

Expand Down Expand Up @@ -508,7 +509,10 @@ def run_orchestrator(
out_path: str,
canary: str | None,
) -> None:
client = anthropic.Anthropic()
client = anthropic.Anthropic(
base_url="https://openrouter.ai/api",
auth_token=os.environ.get("OPENROUTER_API_KEY"),
)

transcript_events: list[dict] = []
audit_entries: list[dict] = []
Expand Down