Move to GPT for API consistency

drb-c2-core/.env.example

@@ -18,11 +18,8 @@ GCS_BUCKET=your-bucket-name
 # How long (seconds) before a node is marked offline if no checkin received
 NODE_OFFLINE_THRESHOLD=90
 
-# OpenAI Whisper — for audio transcription
+# OpenAI — for transcription (Whisper), intelligence extraction, embeddings, and summaries
 OPENAI_API_KEY=
-
-# Gemini — for intelligence extraction, embeddings, and incident summaries
-GEMINI_API_KEY=
 SUMMARY_INTERVAL_MINUTES=15
 CORRELATION_WINDOW_HOURS=4
 EMBEDDING_SIMILARITY_THRESHOLD=0.82

drb-c2-core/app/internal/intelligence.py

@@ -113,20 +113,14 @@ def _sync_extract(
     system_id: Optional[str],
     segments: Optional[list[dict]],
 ) -> dict:
-    """Call Gemini Flash and parse the JSON response."""
+    """Call GPT-4o mini and parse the JSON response."""
     from app.config import settings
-    import google.generativeai as genai
+    from openai import OpenAI
 
-    if not settings.gemini_api_key:
-        logger.warning("GEMINI_API_KEY not set — intelligence extraction disabled.")
+    if not settings.openai_api_key:
+        logger.warning("OPENAI_API_KEY not set — intelligence extraction disabled.")
         return {}
 
-    genai.configure(api_key=settings.gemini_api_key)
-    model = genai.GenerativeModel(
-        "gemini-2.5-flash-lite",
-        generation_config={"response_mime_type": "application/json"},
-    )
-
     tg = f"{talkgroup_name} (TGID {talkgroup_id})" if talkgroup_id else (talkgroup_name or "unknown")
     prompt = _PROMPT_TEMPLATE.format(
         transcript_block=_build_transcript_block(transcript, segments),
@@ -135,13 +129,18 @@ def _sync_extract(
     )
 
     try:
-        response = model.generate_content(prompt)
-        return json.loads(response.text)
+        client = OpenAI(api_key=settings.openai_api_key)
+        response = client.chat.completions.create(
+            model="gpt-4o-mini",
+            messages=[{"role": "user", "content": prompt}],
+            response_format={"type": "json_object"},
+        )
+        return json.loads(response.choices[0].message.content)
     except json.JSONDecodeError as e:
-        logger.warning(f"Gemini returned non-JSON: {e}")
+        logger.warning(f"GPT-4o mini returned non-JSON: {e}")
         return {}
     except Exception as e:
-        logger.warning(f"Gemini extraction failed: {e}")
+        logger.warning(f"GPT-4o mini extraction failed: {e}")
         return {}
 
 

drb-c2-core/app/internal/summarizer.py

@@ -76,14 +76,11 @@ async def _summarize_incident(inc: dict) -> None:
 
 def _sync_summarize(inc: dict, transcripts: list[str]) -> Optional[str]:
     from app.config import settings
-    import google.generativeai as genai
+    from openai import OpenAI
 
-    if not settings.gemini_api_key:
+    if not settings.openai_api_key:
         return None
 
-    genai.configure(api_key=settings.gemini_api_key)
-    model = genai.GenerativeModel("gemini-2.5-flash-lite")
-
     inc_type = inc.get("type", "unknown")
     location = inc.get("location") or "unknown location"
     tg_ids = ", ".join(inc.get("talkgroup_ids", [])) or "unknown"
@@ -107,8 +104,12 @@ Write a concise factual summary of this incident in 2-4 sentences. Include:
 Be factual. Do not speculate beyond what the transcripts say. Do not use bullet points."""
 
     try:
-        response = model.generate_content(prompt)
-        return response.text.strip() or None
+        client = OpenAI(api_key=settings.openai_api_key)
+        response = client.chat.completions.create(
+            model="gpt-4o-mini",
+            messages=[{"role": "user", "content": prompt}],
+        )
+        return response.choices[0].message.content.strip() or None
     except Exception as e:
-        logger.warning(f"Gemini summary failed: {e}")
+        logger.warning(f"GPT-4o mini summary failed: {e}")
         return None