diff --git a/app.py b/app.py index c391eda..4fe9860 100644 --- a/app.py +++ b/app.py @@ -6,14 +6,14 @@ from typing import Optional import requests from fastapi import FastAPI, File, Form, HTTPException, UploadFile -from fastapi.responses import HTMLResponse +from fastapi.responses import HTMLResponse, PlainTextResponse API_BASE = os.getenv("API_BASE", "http://gx10.aquantico.lan:8093").rstrip("/") OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://gx10.aquantico.lan:11434").rstrip("/") OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "qwen3.5:9b") DB_PATH = os.getenv("DB_PATH", "/data/ui.db") -app = FastAPI(title="Diarization UI + LLM") +app = FastAPI(title="Diarization UI") def db(): @@ -22,32 +22,114 @@ def db(): return conn +def now_iso() -> str: + return datetime.utcnow().isoformat() + + def init_db(): os.makedirs(os.path.dirname(DB_PATH), exist_ok=True) with db() as c: c.execute( """ - CREATE TABLE IF NOT EXISTS transcripts ( + CREATE TABLE IF NOT EXISTS projects ( id INTEGER PRIMARY KEY AUTOINCREMENT, - created_at TEXT NOT NULL, - filename TEXT, - formatted_text TEXT NOT NULL, - raw_json TEXT NOT NULL + name TEXT UNIQUE NOT NULL, + created_at TEXT NOT NULL ) """ ) c.execute( """ - CREATE TABLE IF NOT EXISTS analyses ( + CREATE TABLE IF NOT EXISTS prompts ( id INTEGER PRIMARY KEY AUTOINCREMENT, - transcript_id INTEGER NOT NULL, - created_at TEXT NOT NULL, + name TEXT UNIQUE NOT NULL, prompt TEXT NOT NULL, - answer TEXT NOT NULL, - FOREIGN KEY(transcript_id) REFERENCES transcripts(id) + created_at TEXT NOT NULL, + updated_at TEXT NOT NULL ) """ ) + c.execute( + """ + CREATE TABLE IF NOT EXISTS documents ( + id INTEGER PRIMARY KEY AUTOINCREMENT, + project_id INTEGER NOT NULL, + kind TEXT NOT NULL, -- transcript|analysis + title TEXT NOT NULL, + content_md TEXT NOT NULL, + source_document_id INTEGER, + prompt_id INTEGER, + raw_json TEXT, + created_at TEXT NOT NULL, + FOREIGN KEY(project_id) REFERENCES projects(id), + FOREIGN KEY(source_document_id) REFERENCES documents(id), + FOREIGN KEY(prompt_id) REFERENCES prompts(id) + ) + """ + ) + + # defaults + c.execute("INSERT OR IGNORE INTO projects(name, created_at) VALUES (?,?)", ("Default", now_iso())) + c.execute( + "INSERT OR IGNORE INTO prompts(name, prompt, created_at, updated_at) VALUES (?,?,?,?)", + ( + "Zusammenfassung", + "Erstelle eine prägnante Zusammenfassung des Gesprächs in Stichpunkten.", + now_iso(), + now_iso(), + ), + ) + c.execute( + "INSERT OR IGNORE INTO prompts(name, prompt, created_at, updated_at) VALUES (?,?,?,?)", + ( + "Aufgaben", + "Extrahiere alle Aufgaben. Gib pro Aufgabe: Verantwortlich, Aufgabe, Deadline (falls vorhanden), Priorität.", + now_iso(), + now_iso(), + ), + ) + + +def layout(title: str, body: str) -> str: + return f""" + + +{title} + + + +
{body}
+ +""" + + +def get_projects(): + with db() as c: + return c.execute("SELECT id,name FROM projects ORDER BY name").fetchall() + + +def get_prompts(): + with db() as c: + return c.execute("SELECT id,name,prompt FROM prompts ORDER BY name").fetchall() @app.on_event("startup") @@ -57,155 +139,259 @@ def startup(): @app.get("/healthz") def healthz(): - return { - "ok": True, - "api_base": API_BASE, - "ollama_base_url": OLLAMA_BASE_URL, - "ollama_model": OLLAMA_MODEL, - "db_path": DB_PATH, - } + return {"ok": True, "api_base": API_BASE, "ollama_base_url": OLLAMA_BASE_URL, "ollama_model": OLLAMA_MODEL, "db_path": DB_PATH} @app.get("/", response_class=HTMLResponse) -def index(): - return """ - - -Diarization UI - - - -

Upload -> Transcribe + Diarize -> speichern -> LLM Analyse

-
- - -
-

-

-
-
Analyse

-

-
-
-
-

-

-
-
Gespeicherte Transkripte

-
-

-
-
-
+def upload_page(msg: str = ""):
+    projects = get_projects()
+    opts = "".join([f"" for p in projects])
+    body = f"""
+
Audio Upload

+
Audio wird transkribiert + mit Sprechern angereichert und als Dokument gespeichert.

+{f"
{msg}
" if msg else ""}
+

+  

+    
+    
+    
+  

+  

+    
+    
+  

+

 """
+    return layout("Upload", body)
 
 
-@app.post("/process")
-async def process(file: UploadFile = File(...)):
+@app.post("/projects", response_class=HTMLResponse)
+def add_project(name: str = Form(...)):
+    with db() as c:
+        c.execute("INSERT INTO projects(name, created_at) VALUES (?,?)", (name.strip(), now_iso()))
+    return HTMLResponse("")
+
+
+@app.post("/upload", response_class=HTMLResponse)
+async def upload(project_id: int = Form(...), title: str = Form(""), file: UploadFile = File(...)):
     data = await file.read()
     if not data:
-        raise HTTPException(400, "empty file")
+        raise HTTPException(400, "Leere Datei")
 
     files = {"file": (file.filename or "audio.bin", data, file.content_type or "application/octet-stream")}
-    try:
-        r = requests.post(f"{API_BASE}/transcribe-diarize", files=files, timeout=1800)
-    except Exception as e:
-        raise HTTPException(502, f"API unreachable: {e}")
-
+    r = requests.post(f"{API_BASE}/transcribe-diarize", files=files, timeout=1800)
     if r.status_code >= 400:
         raise HTTPException(r.status_code, r.text)
 
     payload = r.json()
-    formatted = payload.get("formatted_text", "")
+    content_md = payload.get("formatted_text", "")
+    doc_title = (title or "").strip() or (file.filename or "Transkript")
 
     with db() as c:
         cur = c.execute(
-            "INSERT INTO transcripts(created_at, filename, formatted_text, raw_json) VALUES (?,?,?,?)",
-            (datetime.utcnow().isoformat(), file.filename, formatted, json.dumps(payload, ensure_ascii=False)),
+            "INSERT INTO documents(project_id, kind, title, content_md, raw_json, created_at) VALUES (?,?,?,?,?,?)",
+            (project_id, "transcript", doc_title, content_md, json.dumps(payload, ensure_ascii=False), now_iso()),
         )
-        transcript_id = cur.lastrowid
+        doc_id = cur.lastrowid
 
-    return {"ok": True, "transcript_id": transcript_id, **payload}
+    return HTMLResponse(f"")
 
 
-@app.get("/transcripts")
-def transcripts(limit: int = 20):
+@app.get("/library", response_class=HTMLResponse)
+def library(project_id: Optional[int] = None):
     with db() as c:
-        rows = c.execute(
-            "SELECT id, created_at, filename, formatted_text FROM transcripts ORDER BY id DESC LIMIT ?",
-            (limit,),
-        ).fetchall()
-    return {"items": [dict(r) for r in rows]}
+        projects = c.execute("SELECT id,name FROM projects ORDER BY name").fetchall()
+        if project_id:
+            docs = c.execute(
+                """
+                SELECT d.id,d.kind,d.title,d.created_at,p.name AS project
+                FROM documents d JOIN projects p ON p.id=d.project_id
+                WHERE d.project_id=? ORDER BY d.id DESC
+                """,
+                (project_id,),
+            ).fetchall()
+        else:
+            docs = c.execute(
+                """
+                SELECT d.id,d.kind,d.title,d.created_at,p.name AS project
+                FROM documents d JOIN projects p ON p.id=d.project_id
+                ORDER BY d.id DESC LIMIT 200
+                """
+            ).fetchall()
+
+    p_opts = "" + "".join(
+        [f"" for p in projects]
+    )
+    items = "".join(
+        [
+            f"
#{d['id']} [{d['kind']}] {d['title']}
{d['project']} · {d['created_at']}
"
+            f"Ansehen | Download .md
"
+            for d in docs
+        ]
+    )
+    body = f"""
+
Datenbank / Dokumente

+

+
+
+
+

+{items or '
Keine Einträge.
'}
+"""
+    return layout("Library", body)
 
 
-@app.post("/analyze")
-def analyze(transcript_id: int = Form(...), prompt: str = Form(...)):
+@app.get("/document/{doc_id}", response_class=HTMLResponse)
+def view_document(doc_id: int):
     with db() as c:
-        row = c.execute("SELECT formatted_text FROM transcripts WHERE id=?", (transcript_id,)).fetchone()
-    if not row:
-        raise HTTPException(404, "transcript not found")
+        d = c.execute(
+            """
+            SELECT d.*, p.name AS project, pr.name AS prompt_name
+            FROM documents d
+            JOIN projects p ON p.id=d.project_id
+            LEFT JOIN prompts pr ON pr.id=d.prompt_id
+            WHERE d.id=?
+            """,
+            (doc_id,),
+        ).fetchone()
+    if not d:
+        raise HTTPException(404, "not found")
+
+    body = f"""
+
Dokument #{d['id']} – {d['title']}

+
Projekt: {d['project']} · Typ: {d['kind']} · {d['created_at']}

+
Download .md

+
{(d['content_md'] or '').replace('<','<')}

+"""
+    return layout("Dokument", body)
+
+
+@app.get("/document/{doc_id}.md", response_class=PlainTextResponse)
+def download_md(doc_id: int):
+    with db() as c:
+        d = c.execute("SELECT title,content_md FROM documents WHERE id=?", (doc_id,)).fetchone()
+    if not d:
+        raise HTTPException(404, "not found")
+    return PlainTextResponse(d["content_md"], headers={"Content-Disposition": f"attachment; filename=document_{doc_id}.md"})
+
+
+@app.get("/prompts", response_class=HTMLResponse)
+def prompts_page():
+    with db() as c:
+        prompts = c.execute("SELECT * FROM prompts ORDER BY name").fetchall()
+        projects = c.execute("SELECT id,name FROM projects ORDER BY name").fetchall()
+
+    p_list = "".join(
+        [
+            f"
{p['name']}
{(p['prompt'] or '').replace('<','<')}
"
+            f"


"
+            for p in prompts
+        ]
+    )
+    project_opts = "".join([f"" for p in projects])
+
+    body = f"""
+
Prompt-Konfiguration

+

+

+  
Neuer Prompt

+  
+  

+  

+

+

+

+

+  
Neues Projekt

+  
+  {project_opts}
+  
+

+

+{p_list}
+"""
+    return layout("Prompts", body)
+
+
+@app.post("/prompts/add", response_class=HTMLResponse)
+def prompt_add(name: str = Form(...), prompt: str = Form(...)):
+    with db() as c:
+        c.execute(
+            "INSERT INTO prompts(name,prompt,created_at,updated_at) VALUES (?,?,?,?)",
+            (name.strip(), prompt.strip(), now_iso(), now_iso()),
+        )
+    return HTMLResponse("")
+
+
+@app.post("/prompts/update", response_class=HTMLResponse)
+def prompt_update(id: int = Form(...), name: str = Form(...), prompt: str = Form(...)):
+    with db() as c:
+        c.execute("UPDATE prompts SET name=?, prompt=?, updated_at=? WHERE id=?", (name.strip(), prompt.strip(), now_iso(), id))
+    return HTMLResponse("")
+
+
+@app.get("/run", response_class=HTMLResponse)
+def run_page():
+    with db() as c:
+        docs = c.execute("SELECT id,title,kind,created_at FROM documents ORDER BY id DESC LIMIT 200").fetchall()
+        prompts = c.execute("SELECT id,name FROM prompts ORDER BY name").fetchall()
+
+    d_opts = "".join([f"" for d in docs])
+    p_opts = "".join([f"" for p in prompts])
+
+    body = f"""
+
Prompt ausführen

+

+  

+  


+  

+  


+  
+

+"""
+    return layout("Run", body)
+
+
+@app.post("/run", response_class=HTMLResponse)
+def run_prompt(document_id: int = Form(...), prompt_id: int = Form(...)):
+    with db() as c:
+        doc = c.execute("SELECT * FROM documents WHERE id=?", (document_id,)).fetchone()
+        prm = c.execute("SELECT * FROM prompts WHERE id=?", (prompt_id,)).fetchone()
+    if not doc or not prm:
+        raise HTTPException(404, "Dokument oder Prompt nicht gefunden")
 
-    transcript_text = row[0]
     llm_prompt = (
-        "Du bist ein Meeting-Analyst. Arbeite auf Deutsch.\n"
-        "Erzeuge präzise Ausgabe für den folgenden Auftrag.\n\n"
-        f"AUFTRAG:\n{prompt}\n\n"
-        f"TRANSKRIPT:\n{transcript_text}\n"
+        "Du bist ein präziser Assistent. Antworte auf Deutsch.\n"
+        f"AUFTRAG:\n{prm['prompt']}\n\n"
+        f"TEXT:\n{doc['content_md']}\n"
     )
 
-    body = {
-        "model": OLLAMA_MODEL,
-        "prompt": llm_prompt,
-        "stream": False,
-    }
-    try:
-        r = requests.post(f"{OLLAMA_BASE_URL}/api/generate", json=body, timeout=600)
-    except Exception as e:
-        raise HTTPException(502, f"Ollama unreachable: {e}")
-
+    r = requests.post(
+        f"{OLLAMA_BASE_URL}/api/generate",
+        json={"model": OLLAMA_MODEL, "prompt": llm_prompt, "stream": False},
+        timeout=1200,
+    )
     if r.status_code >= 400:
         raise HTTPException(r.status_code, r.text)
-
-    j = r.json()
-    answer = j.get("response", "")
+    answer = r.json().get("response", "")
 
     with db() as c:
         cur = c.execute(
-            "INSERT INTO analyses(transcript_id, created_at, prompt, answer) VALUES (?,?,?,?)",
-            (transcript_id, datetime.utcnow().isoformat(), prompt, answer),
+            """
+            INSERT INTO documents(project_id, kind, title, content_md, source_document_id, prompt_id, raw_json, created_at)
+            VALUES (?,?,?,?,?,?,?,?)
+            """,
+            (
+                doc["project_id"],
+                "analysis",
+                f"Analyse: {prm['name']} · {doc['title']}",
+                answer,
+                doc["id"],
+                prm["id"],
+                json.dumps({"ollama_response": r.json()}, ensure_ascii=False),
+                now_iso(),
+            ),
         )
-        analysis_id = cur.lastrowid
+        new_id = cur.lastrowid
 
-    return {"ok": True, "analysis_id": analysis_id, "answer": answer}
+    return HTMLResponse(f"")