Update app.py
Browse files
app.py
CHANGED
|
@@ -30,7 +30,7 @@ HF_TOKEN = os.getenv("HF_TOKEN")
|
|
| 30 |
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
|
| 31 |
GEMINI_MODEL = os.getenv("GEMINI_MODEL", "gemini-3.8-flash")
|
| 32 |
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
|
| 33 |
-
GROQ_MODEL = os.getenv("GROQ_MODEL"
|
| 34 |
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")
|
| 35 |
OPENROUTER_MODEL = os.getenv("OPENROUTER_MODEL", "meta-llama/llama-3.3-70b-instruct:free")
|
| 36 |
ANY_KEY = bool(GROQ_API_KEY or OPENROUTER_API_KEY or GEMINI_API_KEY or HF_TOKEN)
|
|
@@ -160,10 +160,33 @@ def youtube_transcript(url: str) -> str:
|
|
| 160 |
|
| 161 |
|
| 162 |
# ----------------------------------------------------------------- AGENT
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163 |
def build_model():
|
| 164 |
if GROQ_API_KEY:
|
| 165 |
return LiteLLMModel(
|
| 166 |
-
model_id=f"groq/{
|
| 167 |
api_key=GROQ_API_KEY,
|
| 168 |
timeout=60,
|
| 169 |
num_retries=3,
|
|
@@ -309,7 +332,8 @@ def test_one():
|
|
| 309 |
try:
|
| 310 |
agent = BasicAgent(max_steps=3)
|
| 311 |
q = ".rewsna eht sa \"tfel\" drow eht fo etisoppo eht etirw ,ecnetnes siht dnatsrednu uoy fI"
|
| 312 |
-
|
|
|
|
| 313 |
except Exception as e:
|
| 314 |
return f"FAILED: {type(e).__name__}: {e}"
|
| 315 |
|
|
|
|
| 30 |
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
|
| 31 |
GEMINI_MODEL = os.getenv("GEMINI_MODEL", "gemini-3.8-flash")
|
| 32 |
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
|
| 33 |
+
GROQ_MODEL = os.getenv("GROQ_MODEL") # optional override
|
| 34 |
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")
|
| 35 |
OPENROUTER_MODEL = os.getenv("OPENROUTER_MODEL", "meta-llama/llama-3.3-70b-instruct:free")
|
| 36 |
ANY_KEY = bool(GROQ_API_KEY or OPENROUTER_API_KEY or GEMINI_API_KEY or HF_TOKEN)
|
|
|
|
| 160 |
|
| 161 |
|
| 162 |
# ----------------------------------------------------------------- AGENT
|
| 163 |
+
def pick_groq_model() -> str:
|
| 164 |
+
"""Ask Groq which models this key can use and pick the strongest chat model."""
|
| 165 |
+
if GROQ_MODEL:
|
| 166 |
+
return GROQ_MODEL
|
| 167 |
+
r = requests.get(
|
| 168 |
+
"https://api.groq.com/openai/v1/models",
|
| 169 |
+
headers={"Authorization": f"Bearer {GROQ_API_KEY}"},
|
| 170 |
+
timeout=30,
|
| 171 |
+
)
|
| 172 |
+
r.raise_for_status()
|
| 173 |
+
ids = [m["id"] for m in r.json().get("data", [])]
|
| 174 |
+
skip = ("whisper", "guard", "tts", "orpheus", "embed", "prompt-guard", "safeguard", "playai", "distil")
|
| 175 |
+
chat = [i for i in ids if not any(s in i.lower() for s in skip)]
|
| 176 |
+
prefs = ("gpt-oss-120b", "llama-3.3-70b", "70b", "qwen", "kimi", "gpt-oss-20b", "llama")
|
| 177 |
+
for p in prefs:
|
| 178 |
+
for i in chat:
|
| 179 |
+
if p in i.lower():
|
| 180 |
+
return i
|
| 181 |
+
if chat:
|
| 182 |
+
return chat[0]
|
| 183 |
+
raise RuntimeError(f"No usable chat model found. Models seen: {ids}")
|
| 184 |
+
|
| 185 |
+
|
| 186 |
def build_model():
|
| 187 |
if GROQ_API_KEY:
|
| 188 |
return LiteLLMModel(
|
| 189 |
+
model_id=f"groq/{pick_groq_model()}",
|
| 190 |
api_key=GROQ_API_KEY,
|
| 191 |
timeout=60,
|
| 192 |
num_retries=3,
|
|
|
|
| 332 |
try:
|
| 333 |
agent = BasicAgent(max_steps=3)
|
| 334 |
q = ".rewsna eht sa \"tfel\" drow eht fo etisoppo eht etirw ,ecnetnes siht dnatsrednu uoy fI"
|
| 335 |
+
used = pick_groq_model() if GROQ_API_KEY else "n/a"
|
| 336 |
+
return f"Model: {used}\nAnswer: {agent(q, 'test', '')} (expected: right)"
|
| 337 |
except Exception as e:
|
| 338 |
return f"FAILED: {type(e).__name__}: {e}"
|
| 339 |
|