Rename llm to prompt, add LinkedIn cache

Migrate storage and API from 'llm' to 'prompt': update SQL create table, change all queries/inserts to use prompt, remove tsvector/search_vector usage, and update logging/meta messages. Add drop_llm_table.sql to clean up old table. Enhance POST /prompt/linkedin to check the prompt table for a cached completion by linkedinUrl (returns cached result or a warning if not found). Update prospects endpoint to fetch related prompt_records instead of llm_records and adjust returned fields. Minor docstring and metadata text updates.

Author: goldlabelapps

app/api/prompt/__init__.py

@@ -1,4 +1,4 @@
-"""LLM Routes"""
+"""Prompt Routes"""
 
 from .prompt import router as prompt_router
 from .linkedin import router as linkedin_router

app/api/prompt/linkedin.py

@@ -1,15 +1,67 @@
-from fastapi import APIRouter, Depends
+from fastapi import APIRouter, Depends, HTTPException
 
 from app.utils.api_key_auth import get_api_key
+from app.utils.db import get_db_connection_direct
 from app.utils.make_meta import make_meta
 
 router = APIRouter()
 
 
 @router.post("/prompt/linkedin")
-def linkedin_prompt_success(api_key: str = Depends(get_api_key)) -> dict:
-    """POST /prompt/linkedin: Success stub endpoint."""
-    return {
-        "meta": make_meta("success", "LinkedIn prompt endpoint working"),
-        "message": "LinkedIn prompt endpoint is live.",
-    }
+def linkedin_prompt_success(payload: dict, api_key: str = Depends(get_api_key)) -> dict:
+    """POST /prompt/linkedin: return cached completion for linkedinUrl when available."""
+    linkedin_url = (payload.get("linkedinUrl") or "").strip()
+    if not linkedin_url:
+        raise HTTPException(status_code=400, detail="Missing 'linkedinUrl' in request body.")
+
+    conn = None
+    cur = None
+    try:
+        conn = get_db_connection_direct()
+        cur = conn.cursor()
+        cur.execute(
+            """
+            SELECT id, completion, time, model, data
+            FROM prompt
+            WHERE (data->>'linkedinUrl' = %s OR prompt ILIKE %s)
+            ORDER BY id DESC
+            LIMIT 1;
+            """,
+            (linkedin_url, f"%{linkedin_url}%"),
+        )
+        row = cur.fetchone()
+
+        if row:
+            return {
+                "meta": make_meta("success", "LinkedIn URL already analysed"),
+                "data": {
+                    "cached": True,
+                    "id": row[0],
+                    "linkedinUrl": linkedin_url,
+                    "completion": row[1],
+                    "time": row[2].isoformat() if row[2] else None,
+                    "model": row[3],
+                    "record_data": row[4],
+                },
+            }
+
+        return {
+            "meta": make_meta("warning", "LinkedIn URL not analysed yet"),
+            "data": {
+                "cached": False,
+                "linkedinUrl": linkedin_url,
+                "completion": None,
+            },
+        }
+    except HTTPException:
+        raise
+    except Exception as e:
+        return {
+            "meta": make_meta("error", f"DB error: {str(e)}"),
+            "data": {},
+        }
+    finally:
+        if cur:
+            cur.close()
+        if conn:
+            conn.close()

app/api/prompt/prompt.py

@@ -21,8 +21,8 @@ def get_prompt_records(
         if prospect_id is not None:
             # No pagination for single prospect_id lookup
             select_query = """
-                SELECT id, prompt, completion, duration, time, data, model, prospect_id, search_vector
-                FROM llm
+                SELECT id, prompt, completion, duration, time, data, model, prospect_id
+                FROM prompt
                 WHERE prospect_id = %s
                 ORDER BY id DESC
             """
@@ -38,7 +38,6 @@ def get_prompt_records(
                     "data": row[5],
                     "model": row[6],
                     "prospect_id": row[7],
-                    "search_vector": str(row[8]) if row[8] is not None else None,
                 }
                 for row in rows
             ]
@@ -58,12 +57,12 @@ def get_prompt_records(
                 }
         else:
             offset = (page - 1) * page_size
-            cur.execute("SELECT COUNT(*) FROM llm;")
+            cur.execute("SELECT COUNT(*) FROM prompt;")
             count_row = cur.fetchone()
             total = count_row[0] if count_row and count_row[0] is not None else 0
             cur.execute("""
-                SELECT id, prompt, completion, duration, time, data, model, prospect_id, search_vector
-                FROM llm
+                SELECT id, prompt, completion, duration, time, data, model, prospect_id
+                FROM prompt
                 ORDER BY id DESC
                 LIMIT %s OFFSET %s;
             """, (page_size, offset))
@@ -77,13 +76,12 @@ def get_prompt_records(
                     "data": row[5],
                     "model": row[6],
                     "prospect_id": row[7],
-                    "search_vector": str(row[8]) if row[8] is not None else None,
                 }
                 for row in cur.fetchall()
             ]
             cur.close()
             conn.close()
-            meta = make_meta("success", f"LLM {len(records)} records (page {page})")
+            meta = make_meta("success", f"Prompt {len(records)} records (page {page})")
             return {
                 "meta": meta,
                 "data": {
@@ -140,22 +138,21 @@ def llm_post(payload: dict) -> dict:
         if not completion:
             error_details = " | ".join([f"{k}: {v}" for k, v in errors.items()])
             raise Exception(f"No available Gemini model succeeded for generate_content with your API key. Details: {error_details}")
-        # Insert record into llm table
+        # Insert record into prompt table
         record_id = None
         try:
             import json
             from app import __version__
             data_blob = json.dumps({"version": __version__})
             conn = get_db_connection_direct()
             cur = conn.cursor()
-            # Generate tsvector from prompt and completion
             cur.execute(
                 """
-                INSERT INTO llm (prompt, completion, duration, data, model, prospect_id, search_vector)
-                VALUES (%s, %s, %s, %s, %s, %s, to_tsvector('english', %s || ' ' || %s))
+                INSERT INTO prompt (prompt, completion, duration, data, model, prospect_id)
+                VALUES (%s, %s, %s, %s, %s, %s)
                 RETURNING id;
                 """,
-                (prompt, completion, duration, data_blob, used_model, prospect_id, prompt, completion)
+                (prompt, completion, duration, data_blob, used_model, prospect_id)
             )
             record_id_row = cur.fetchone()
             record_id = record_id_row[0] if record_id_row else None
@@ -164,7 +161,7 @@ def llm_post(payload: dict) -> dict:
             conn.close()
         except Exception as db_exc:
             # Log DB error but do not fail the API response
-            logging.error(f"Failed to insert llm record: {db_exc}")
+            logging.error(f"Failed to insert prompt record: {db_exc}")
         meta = make_meta("success", f"Gemini completion received from {used_model}")
         return {"meta": meta, "data": {"id": record_id, "prompt": prompt, "completion": completion}}
     except Exception as e:

app/api/prompt/sql/create_table.sql

@@ -1,5 +1,5 @@
 
-CREATE TABLE IF NOT EXISTS llm (
+CREATE TABLE IF NOT EXISTS prompt (
 	id SERIAL PRIMARY KEY,
 	vector vector(1536),
 	prompt TEXT NOT NULL,

app/api/prompt/sql/drop_llm_table.sql

@@ -0,0 +1 @@
+DROP TABLE IF EXISTS llm;

app/api/prospects/prospects.py

@@ -98,28 +98,27 @@ def prospects_read_one(id: int = Path(..., description="ID of the prospect to re
             if row is not None:
                 columns = [desc[0] for desc in cur.description]
                 data = dict(zip(columns, row))
-                # Fetch related llm records
+                # Fetch related prompt records
                 try:
                     from app.utils.db import get_db_connection_direct
-                    llm_conn = get_db_connection_direct()
-                    llm_cur = llm_conn.cursor()
-                    llm_cur.execute("SELECT id, duration, time, data, model, search_vector FROM llm WHERE prospect_id = %s ORDER BY id DESC;", (id,))
-                    llm_records = [
+                    prompt_conn = get_db_connection_direct()
+                    prompt_cur = prompt_conn.cursor()
+                    prompt_cur.execute("SELECT id, duration, time, data, model FROM prompt WHERE prospect_id = %s ORDER BY id DESC;", (id,))
+                    prompt_records = [
                         {
                             "id": r[0],
                             "duration": r[1],
                             "time": r[2].isoformat() if r[2] else None,
                             "data": r[3],
                             "model": r[4],
-                            "search_vector": str(r[5]) if r[5] is not None else None,
                         }
-                        for r in llm_cur.fetchall()
+                        for r in prompt_cur.fetchall()
                     ]
-                    llm_cur.close()
-                    llm_conn.close()
-                    data["llm_records"] = llm_records
-                except Exception as llm_exc:
-                    data["llm_records"] = []
+                    prompt_cur.close()
+                    prompt_conn.close()
+                    data["prompt_records"] = prompt_records
+                except Exception as prompt_exc:
+                    data["prompt_records"] = []
             else:
                 data = None
                 meta = make_meta("error", f"No prospect found with id {id}")