Spaces:
Sleeping
Sleeping
Commit ·
112b4f4
1
Parent(s): b71f629
fix: curated model lists, reasoning model token scaling, skip temp files
Browse files- Restrict each provider to 3 supported models instead of fetching all
from API (OpenAI: gpt-4.1/4o/4o-mini, Anthropic: sonnet/opus/haiku,
Gemini: 2.5-pro/2.5-flash/2.0-flash)
- Scale up max_completion_tokens 4x for OpenAI reasoning models (o1/o3/o4)
since reasoning tokens count against the budget
- Skip ~$ temp Excel files during ingestion
- Log LLM errors in codebook description generation for debugging
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- setup.py +1 -1
- src/ingest.py +1 -1
- src/llm/anthropic.py +2 -8
- src/llm/gemini.py +2 -12
- src/llm/openai.py +6 -10
- src/sql_ingest.py +2 -1
setup.py
CHANGED
|
@@ -144,7 +144,7 @@ def run_wizard():
|
|
| 144 |
except Exception:
|
| 145 |
print(f" Warning: Could not validate API key for {provider}. Using default model list.")
|
| 146 |
fallback = {
|
| 147 |
-
"openai": ["gpt-
|
| 148 |
"anthropic": ["claude-sonnet-4-6", "claude-opus-4-6", "claude-haiku-4-5"],
|
| 149 |
"gemini": ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.0-flash"],
|
| 150 |
}
|
|
|
|
| 144 |
except Exception:
|
| 145 |
print(f" Warning: Could not validate API key for {provider}. Using default model list.")
|
| 146 |
fallback = {
|
| 147 |
+
"openai": ["gpt-4.1", "gpt-4o", "gpt-4o-mini"],
|
| 148 |
"anthropic": ["claude-sonnet-4-6", "claude-opus-4-6", "claude-haiku-4-5"],
|
| 149 |
"gemini": ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.0-flash"],
|
| 150 |
}
|
src/ingest.py
CHANGED
|
@@ -106,7 +106,7 @@ def discover_files(documents_dir: str) -> list[tuple[Path, str]]:
|
|
| 106 |
continue
|
| 107 |
if file_path.suffix.lower() not in supported:
|
| 108 |
continue
|
| 109 |
-
if file_path.name.startswith("."):
|
| 110 |
continue
|
| 111 |
|
| 112 |
rel = file_path.relative_to(root)
|
|
|
|
| 106 |
continue
|
| 107 |
if file_path.suffix.lower() not in supported:
|
| 108 |
continue
|
| 109 |
+
if file_path.name.startswith(".") or file_path.name.startswith("~$"):
|
| 110 |
continue
|
| 111 |
|
| 112 |
rel = file_path.relative_to(root)
|
src/llm/anthropic.py
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
"""Anthropic LLM provider."""
|
| 2 |
-
|
| 3 |
|
| 4 |
def generate(system_prompt: str, user_message: str, api_key: str,
|
| 5 |
model: str = None, temperature: float = 0.0,
|
|
@@ -21,10 +21,4 @@ def generate(system_prompt: str, user_message: str, api_key: str,
|
|
| 21 |
raise RuntimeError(f"Anthropic API error ({error_type}). Check your API key and network connection.") from e
|
| 22 |
|
| 23 |
def list_models(api_key: str) -> list[str]:
|
| 24 |
-
|
| 25 |
-
from anthropic import Anthropic
|
| 26 |
-
client = Anthropic(api_key=api_key)
|
| 27 |
-
models = client.models.list()
|
| 28 |
-
return sorted([m.id for m in models.data])
|
| 29 |
-
except Exception:
|
| 30 |
-
return FALLBACK_MODELS
|
|
|
|
| 1 |
"""Anthropic LLM provider."""
|
| 2 |
+
SUPPORTED_MODELS = ["claude-sonnet-4-6", "claude-opus-4-6", "claude-haiku-4-5"]
|
| 3 |
|
| 4 |
def generate(system_prompt: str, user_message: str, api_key: str,
|
| 5 |
model: str = None, temperature: float = 0.0,
|
|
|
|
| 21 |
raise RuntimeError(f"Anthropic API error ({error_type}). Check your API key and network connection.") from e
|
| 22 |
|
| 23 |
def list_models(api_key: str) -> list[str]:
|
| 24 |
+
return list(SUPPORTED_MODELS)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
src/llm/gemini.py
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
"""Google Gemini LLM provider."""
|
| 2 |
-
|
| 3 |
|
| 4 |
def generate(system_prompt: str, user_message: str, api_key: str,
|
| 5 |
model: str = None, temperature: float = 0.0,
|
|
@@ -26,14 +26,4 @@ def generate(system_prompt: str, user_message: str, api_key: str,
|
|
| 26 |
return "[Gemini error: request failed]"
|
| 27 |
|
| 28 |
def list_models(api_key: str) -> list[str]:
|
| 29 |
-
|
| 30 |
-
import google.generativeai as genai
|
| 31 |
-
genai.configure(api_key=api_key)
|
| 32 |
-
models = genai.list_models()
|
| 33 |
-
chat_models = sorted([
|
| 34 |
-
m.name.replace("models/", "") for m in models
|
| 35 |
-
if "generateContent" in (m.supported_generation_methods or [])
|
| 36 |
-
])
|
| 37 |
-
return chat_models if chat_models else FALLBACK_MODELS
|
| 38 |
-
except Exception:
|
| 39 |
-
return FALLBACK_MODELS
|
|
|
|
| 1 |
"""Google Gemini LLM provider."""
|
| 2 |
+
SUPPORTED_MODELS = ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.0-flash"]
|
| 3 |
|
| 4 |
def generate(system_prompt: str, user_message: str, api_key: str,
|
| 5 |
model: str = None, temperature: float = 0.0,
|
|
|
|
| 26 |
return "[Gemini error: request failed]"
|
| 27 |
|
| 28 |
def list_models(api_key: str) -> list[str]:
|
| 29 |
+
return list(SUPPORTED_MODELS)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
src/llm/openai.py
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
"""OpenAI LLM provider."""
|
| 2 |
-
|
| 3 |
|
| 4 |
def generate(system_prompt: str, user_message: str, api_key: str,
|
| 5 |
model: str = None, temperature: float = 0.0,
|
|
@@ -10,6 +10,9 @@ def generate(system_prompt: str, user_message: str, api_key: str,
|
|
| 10 |
is_reasoning = model.startswith(("o1", "o3", "o4"))
|
| 11 |
try:
|
| 12 |
if is_reasoning:
|
|
|
|
|
|
|
|
|
|
| 13 |
messages = [
|
| 14 |
{"role": "developer", "content": system_prompt},
|
| 15 |
{"role": "user", "content": user_message},
|
|
@@ -17,7 +20,7 @@ def generate(system_prompt: str, user_message: str, api_key: str,
|
|
| 17 |
response = client.chat.completions.create(
|
| 18 |
model=model,
|
| 19 |
messages=messages,
|
| 20 |
-
max_completion_tokens=
|
| 21 |
)
|
| 22 |
else:
|
| 23 |
messages = [
|
|
@@ -38,11 +41,4 @@ def generate(system_prompt: str, user_message: str, api_key: str,
|
|
| 38 |
raise RuntimeError(f"OpenAI API error ({error_type}). Check your API key and network connection.") from e
|
| 39 |
|
| 40 |
def list_models(api_key: str) -> list[str]:
|
| 41 |
-
|
| 42 |
-
from openai import OpenAI
|
| 43 |
-
client = OpenAI(api_key=api_key)
|
| 44 |
-
models = client.models.list()
|
| 45 |
-
chat_models = sorted([m.id for m in models if "gpt" in m.id or m.id.startswith(("o1", "o3", "o4"))])
|
| 46 |
-
return chat_models if chat_models else FALLBACK_MODELS
|
| 47 |
-
except Exception:
|
| 48 |
-
return FALLBACK_MODELS
|
|
|
|
| 1 |
"""OpenAI LLM provider."""
|
| 2 |
+
SUPPORTED_MODELS = ["gpt-4.1", "gpt-4o", "gpt-4o-mini"]
|
| 3 |
|
| 4 |
def generate(system_prompt: str, user_message: str, api_key: str,
|
| 5 |
model: str = None, temperature: float = 0.0,
|
|
|
|
| 10 |
is_reasoning = model.startswith(("o1", "o3", "o4"))
|
| 11 |
try:
|
| 12 |
if is_reasoning:
|
| 13 |
+
# Reasoning models include thinking tokens in max_completion_tokens,
|
| 14 |
+
# so we need a larger budget to get enough output tokens.
|
| 15 |
+
reasoning_budget = max(max_tokens * 4, 4096)
|
| 16 |
messages = [
|
| 17 |
{"role": "developer", "content": system_prompt},
|
| 18 |
{"role": "user", "content": user_message},
|
|
|
|
| 20 |
response = client.chat.completions.create(
|
| 21 |
model=model,
|
| 22 |
messages=messages,
|
| 23 |
+
max_completion_tokens=reasoning_budget,
|
| 24 |
)
|
| 25 |
else:
|
| 26 |
messages = [
|
|
|
|
| 41 |
raise RuntimeError(f"OpenAI API error ({error_type}). Check your API key and network connection.") from e
|
| 42 |
|
| 43 |
def list_models(api_key: str) -> list[str]:
|
| 44 |
+
return list(SUPPORTED_MODELS)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
src/sql_ingest.py
CHANGED
|
@@ -263,7 +263,8 @@ def _describe_columns_with_llm(
|
|
| 263 |
if start >= 0 and end > start:
|
| 264 |
return json.loads(text[start:end])
|
| 265 |
return {}
|
| 266 |
-
except Exception:
|
|
|
|
| 267 |
return {}
|
| 268 |
|
| 269 |
|
|
|
|
| 263 |
if start >= 0 and end > start:
|
| 264 |
return json.loads(text[start:end])
|
| 265 |
return {}
|
| 266 |
+
except Exception as e:
|
| 267 |
+
print(f" LLM description error: {e}")
|
| 268 |
return {}
|
| 269 |
|
| 270 |
|