26362c7359
- Remove _DEFAULT_KEY from real-fit-engine.py, run-focused-eval.py, test_real_api.py - Replace USE_MOCK env var with DRY_RUN --dry-run CLI flag - Create tests/mocks/ollama_mock.py for isolated mock testing - Remove stale glm-5.1 from DEFAULT_MODELS and docs - Update .env.example with OLLAMA_HOST and OLLAMA_KEY Issues: #123
35 lines
1.3 KiB
Python
35 lines
1.3 KiB
Python
#!/usr/bin/env python3
|
|
import sys, os
|
|
# OLLAMA_KEY must be set via environment variable or .env file — never hardcode
|
|
if not os.environ.get("OLLAMA_KEY"):
|
|
print("[FATAL] OLLAMA_KEY not set. Set it via environment variable or .env file.", file=sys.stderr)
|
|
sys.exit(1)
|
|
os.environ.setdefault("OLLAMA_HOST", "https://ollama.com/v1")
|
|
|
|
sys.path.insert(0, "scripts")
|
|
from real_fit_engine import call_ollama, evaluate_response, init_db, import_from_evolution, generate_prompts
|
|
import sqlite3
|
|
|
|
init_db()
|
|
import_from_evolution()
|
|
generate_prompts()
|
|
|
|
conn = sqlite3.connect("agent-evolution/data/real-fit.db")
|
|
row = conn.execute("SELECT system_prompt, user_prompt, expected_keywords, rubric FROM test_prompts WHERE agent_name = ?", ("code-skeptic",)).fetchone()
|
|
conn.close()
|
|
|
|
if row:
|
|
system, user, expected, rubric = row
|
|
print("=== REAL Ollama: code-skeptic x kimi-k2.6 ===")
|
|
resp, latency, tokens = call_ollama("kimi-k2.6", system, user, expected)
|
|
print(f"Latency: {latency}ms")
|
|
print(f"Tokens: {tokens}")
|
|
print("Response (first 300 chars):")
|
|
print(resp[:300])
|
|
print("\n...")
|
|
ev = evaluate_response(resp, expected, rubric)
|
|
print(f"Score: {ev['total']:.1f}")
|
|
print(f"Explanation: {ev['explanation']}")
|
|
else:
|
|
print("No prompt found for code-skeptic")
|