tegwick
880c1d1374
Add GeminiAdapter calling Google's Generative Language REST API (default model: gemini-2.5-flash). Register "gemini" as third provider in the factory and CLI. Add rate-limit retry with exponential backoff to the pipeline's _call_llm helper. Increase default max_tokens from 2000 to 4096. Process book-1-chapter-05 via Gemini free tier — 1 new entity extracted (necessaries-conveniencies-and-amusements-of-life), 41 existing entities correctly skipped by dedup. Canonical set now at 42 unique entities. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
116 lines
3.8 KiB
Python
116 lines
3.8 KiB
Python
"""
|
|
Google Gemini adapter — calls the Generative Language REST API directly.
|
|
"""
|
|
|
|
import time
|
|
from typing import Optional, Dict, Any
|
|
|
|
from markitect.prompts.execution.llm_adapter import LLMAdapter
|
|
from markitect.prompts.execution.models import RunConfig, LLMResponse
|
|
from markitect.llm.config import resolve_api_key, find_project_root
|
|
from markitect.llm._http import post_json
|
|
from markitect.llm.exceptions import LLMConfigurationError
|
|
|
|
_DEFAULT_MODEL = "gemini-2.5-flash"
|
|
_API_BASE = "https://generativelanguage.googleapis.com/v1beta"
|
|
|
|
|
|
class GeminiAdapter(LLMAdapter):
|
|
"""LLM adapter that calls the Google Generative Language API.
|
|
|
|
Supports the free tier of Gemini models via a Google AI Studio API key.
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
model: Optional[str] = None,
|
|
api_key: Optional[str] = None,
|
|
system_prompt: Optional[str] = None,
|
|
**_kwargs: Any,
|
|
):
|
|
self._model = model or _DEFAULT_MODEL
|
|
self._system_prompt = system_prompt
|
|
|
|
root = find_project_root()
|
|
key_file_paths = [root / "apikey-geminifree.txt"] if root else []
|
|
self._api_key = resolve_api_key(
|
|
explicit=api_key,
|
|
env_var="GEMINI_API_KEY",
|
|
key_file_paths=key_file_paths,
|
|
)
|
|
if not self._api_key:
|
|
raise LLMConfigurationError(
|
|
"No Gemini API key found. Set GEMINI_API_KEY or create "
|
|
"apikey-geminifree.txt in the project root.",
|
|
context={"provider": "gemini"},
|
|
)
|
|
|
|
# ── LLMAdapter interface ────────────────────────────────────────
|
|
|
|
def execute_prompt(self, prompt: str, config: RunConfig) -> LLMResponse:
|
|
model = self._model
|
|
|
|
# Build Gemini request
|
|
contents: list[Dict[str, Any]] = []
|
|
if self._system_prompt:
|
|
contents.append({
|
|
"role": "user",
|
|
"parts": [{"text": self._system_prompt}],
|
|
})
|
|
contents.append({
|
|
"role": "model",
|
|
"parts": [{"text": "Understood."}],
|
|
})
|
|
contents.append({
|
|
"role": "user",
|
|
"parts": [{"text": prompt}],
|
|
})
|
|
|
|
payload: Dict[str, Any] = {
|
|
"contents": contents,
|
|
"generationConfig": {
|
|
"temperature": config.temperature,
|
|
"maxOutputTokens": config.max_tokens,
|
|
},
|
|
}
|
|
|
|
url = f"{_API_BASE}/models/{model}:generateContent?key={self._api_key}"
|
|
|
|
start = time.time()
|
|
data = post_json(url, payload, timeout=config.timeout_seconds)
|
|
latency = time.time() - start
|
|
|
|
# Parse Gemini response
|
|
candidates = data.get("candidates", [])
|
|
if not candidates:
|
|
content = ""
|
|
finish_reason = "error"
|
|
else:
|
|
parts = candidates[0].get("content", {}).get("parts", [])
|
|
content = "".join(p.get("text", "") for p in parts)
|
|
finish_reason = candidates[0].get("finishReason", "STOP").lower()
|
|
|
|
usage_meta = data.get("usageMetadata", {})
|
|
|
|
return LLMResponse(
|
|
content=content,
|
|
model=model,
|
|
usage={
|
|
"prompt_tokens": usage_meta.get("promptTokenCount", 0),
|
|
"completion_tokens": usage_meta.get("candidatesTokenCount", 0),
|
|
"total_tokens": usage_meta.get("totalTokenCount", 0),
|
|
},
|
|
finish_reason=finish_reason,
|
|
metadata={
|
|
"provider": "gemini",
|
|
"latency_seconds": round(latency, 3),
|
|
},
|
|
)
|
|
|
|
def validate_config(self, config: RunConfig) -> bool:
|
|
if not self._api_key:
|
|
return False
|
|
if not (0.0 <= config.temperature <= 2.0):
|
|
return False
|
|
return True
|