Spaces:
Sleeping
Sleeping
Fixes env variables.
Browse files- src/app.py +15 -3
- src/context_pilot_workflow.py +14 -2
src/app.py
CHANGED
|
@@ -96,15 +96,27 @@ CONTEXT_STORE_PATH = Path(__file__).parent / ".context_store.json"
|
|
| 96 |
# LLM Client for generating responses (more capable model)
|
| 97 |
# RESPONSE_LLM: Used for generating actual responses (higher quality)
|
| 98 |
# CONTEXT_LLM: Used for topic detection in workflow (cheaper, configured there)
|
| 99 |
-
RESPONSE_LLM_MODEL = os.getenv("RESPONSE_LLM",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
|
| 101 |
# Threshold for summarizing responses before storing in context (in characters)
|
| 102 |
# Responses longer than this will be summarized to reduce context size
|
| 103 |
SUMMARIZE_THRESHOLD = int(os.getenv("SUMMARIZE_THRESHOLD", "500"))
|
| 104 |
|
| 105 |
llm_client = OpenAI(
|
| 106 |
-
api_key=
|
| 107 |
-
base_url=
|
| 108 |
)
|
| 109 |
LLM_MODEL = RESPONSE_LLM_MODEL
|
| 110 |
|
|
|
|
| 96 |
# LLM Client for generating responses (more capable model)
|
| 97 |
# RESPONSE_LLM: Used for generating actual responses (higher quality)
|
| 98 |
# CONTEXT_LLM: Used for topic detection in workflow (cheaper, configured there)
|
| 99 |
+
RESPONSE_LLM_MODEL = os.getenv("RESPONSE_LLM", "openai/gpt-oss-120b")
|
| 100 |
+
NEBIUS_BASE_URL = os.getenv("NEBIUS_BASE_URL")
|
| 101 |
+
NEBIUS_API_KEY = os.getenv("NEBIUS_API_KEY")
|
| 102 |
+
|
| 103 |
+
# Validate required environment variables
|
| 104 |
+
if not NEBIUS_BASE_URL:
|
| 105 |
+
print("WARNING: NEBIUS_BASE_URL not set. Response generation will fail.")
|
| 106 |
+
if not NEBIUS_API_KEY:
|
| 107 |
+
print("WARNING: NEBIUS_API_KEY not set. Response generation will fail.")
|
| 108 |
+
|
| 109 |
+
print(f"[ContextPilot] Response LLM: {RESPONSE_LLM_MODEL}")
|
| 110 |
+
print(f"[ContextPilot] API Base: {NEBIUS_BASE_URL}")
|
| 111 |
+
print(f"[ContextPilot] API Key set: {bool(NEBIUS_API_KEY)}")
|
| 112 |
|
| 113 |
# Threshold for summarizing responses before storing in context (in characters)
|
| 114 |
# Responses longer than this will be summarized to reduce context size
|
| 115 |
SUMMARIZE_THRESHOLD = int(os.getenv("SUMMARIZE_THRESHOLD", "500"))
|
| 116 |
|
| 117 |
llm_client = OpenAI(
|
| 118 |
+
api_key=NEBIUS_API_KEY,
|
| 119 |
+
base_url=NEBIUS_BASE_URL,
|
| 120 |
)
|
| 121 |
LLM_MODEL = RESPONSE_LLM_MODEL
|
| 122 |
|
src/context_pilot_workflow.py
CHANGED
|
@@ -419,11 +419,23 @@ CONTEXT_TOOLS = [
|
|
| 419 |
# Use CONTEXT_LLM for topic detection (cheaper)
|
| 420 |
# Use RESPONSE_LLM for generation (more capable) - configured in app.py
|
| 421 |
CONTEXT_LLM_MODEL = os.getenv("CONTEXT_LLM", os.getenv("NEBIUS_MODEL", "openai/gpt-4o-mini"))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 422 |
|
| 423 |
topic_llm = OpenAILike(
|
| 424 |
model=CONTEXT_LLM_MODEL,
|
| 425 |
-
api_base=
|
| 426 |
-
api_key=
|
| 427 |
is_chat_model=True,
|
| 428 |
is_function_calling_model=True,
|
| 429 |
context_window=128000,
|
|
|
|
| 419 |
# Use CONTEXT_LLM for topic detection (cheaper)
|
| 420 |
# Use RESPONSE_LLM for generation (more capable) - configured in app.py
|
| 421 |
CONTEXT_LLM_MODEL = os.getenv("CONTEXT_LLM", os.getenv("NEBIUS_MODEL", "openai/gpt-4o-mini"))
|
| 422 |
+
NEBIUS_BASE_URL = os.getenv("NEBIUS_BASE_URL")
|
| 423 |
+
NEBIUS_API_KEY = os.getenv("NEBIUS_API_KEY")
|
| 424 |
+
|
| 425 |
+
# Validate required environment variables
|
| 426 |
+
if not NEBIUS_BASE_URL:
|
| 427 |
+
print("WARNING: NEBIUS_BASE_URL not set. Topic detection will fail.")
|
| 428 |
+
if not NEBIUS_API_KEY:
|
| 429 |
+
print("WARNING: NEBIUS_API_KEY not set. Topic detection will fail.")
|
| 430 |
+
|
| 431 |
+
print(f"[ContextPilot] Context LLM: {CONTEXT_LLM_MODEL}")
|
| 432 |
+
print(f"[ContextPilot] API Base: {NEBIUS_BASE_URL}")
|
| 433 |
+
print(f"[ContextPilot] API Key set: {bool(NEBIUS_API_KEY)}")
|
| 434 |
|
| 435 |
topic_llm = OpenAILike(
|
| 436 |
model=CONTEXT_LLM_MODEL,
|
| 437 |
+
api_base=NEBIUS_BASE_URL,
|
| 438 |
+
api_key=NEBIUS_API_KEY,
|
| 439 |
is_chat_model=True,
|
| 440 |
is_function_calling_model=True,
|
| 441 |
context_window=128000,
|