feat: playbook tools, RAG, vision, voice, chat controls, faster boot

- Tool-using playbooks: read-only tool registry (search_memory,
  search_history, search_documents, list_models, get_time), per-playbook
  allowlist, agentic loop, and a "running tool" status indicator.
- Document ingest / RAG: documents table + chunker + embed/cosine retrieval
  reusing the existing stack; Documents page (upload/paste, viewer, delete);
  top-k chunks injected into the system prompt.
- Vision chat: attach images, base64 into /api/chat.
- Voice I/O: Web Speech dictation + read-aloud (browser-native, no backend).
- Chat controls: stop, regenerate, edit-and-resend; num_ctx knob in Settings.
- Per-playbook model override.
- History polish: per-conversation + bulk ShareGPT export.
- Faster ncp start: UI up first, Ollama warms in the background, with a
  per-phase timing readout.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

synapse/ollama_manager.py

@@ -175,7 +175,7 @@ def _preferred_model(models: list, preference) -> str | None:
     return None
 
 
-def _chat_options(temperature: float | None, num_gpu: int | None) -> dict:
+def _chat_options(temperature: float | None, num_gpu: int | None, num_ctx: int | None = None) -> dict:
     """Assemble the Ollama `options` block from the knobs we expose.
 
     Returns an empty dict when nothing is set so callers can omit `options`
@@ -186,6 +186,8 @@ def _chat_options(temperature: float | None, num_gpu: int | None) -> dict:
         opts["temperature"] = temperature
     if num_gpu is not None:
         opts["num_gpu"] = num_gpu
+    if num_ctx:  # 0 / None -> let Ollama use the model default
+        opts["num_ctx"] = num_ctx
     return opts
 
 
@@ -503,10 +505,16 @@ class OllamaManager:
         temperature: float | None = None,
         num_gpu: int | None = None,
         think: bool = False,
+        tools: list | None = None,
+        num_ctx: int | None = None,
         **kwargs,
     ):
         """Multi-turn chat via /api/chat (accepts a messages array with roles).
 
+        When `tools` is given (non-stream only), the request advertises them and
+        the FULL message dict is returned (so the caller sees `tool_calls`);
+        otherwise the response content string is returned as before.
+
         `think` toggles Qwen3-style reasoning. Default off: the hidden <think>
         block is pure latency for chat/memory. Ollama ignores it for models that
         don't support thinking.
@@ -516,12 +524,14 @@ class OllamaManager:
             if stream:
                 return self._chat_stream(
                     messages=messages, model=model, temperature=temperature,
-                    num_gpu=num_gpu, think=think, start=start,
+                    num_gpu=num_gpu, think=think, start=start, num_ctx=num_ctx,
                 )
             else:
                 body: dict = {"model": model, "messages": messages, "stream": False}
                 body["think"] = think
-                opts = _chat_options(temperature, num_gpu)
+                if tools:
+                    body["tools"] = tools
+                opts = _chat_options(temperature, num_gpu, num_ctx)
                 if opts:
                     body["options"] = opts
                 self._apply_keep_alive(body)
@@ -529,7 +539,9 @@ class OllamaManager:
                     r = await client.post(f"{self._api_base}/api/chat", json=body)
                 elapsed = time.perf_counter() - start
                 r.raise_for_status()
-                return r.json().get("message", {}).get("content", "")
+                message = r.json().get("message", {})
+                # Tool callers need the whole message (tool_calls); others want content.
+                return message if tools else message.get("content", "")
         except Exception as e:
             elapsed = time.perf_counter() - start
             _log.exception("chat error after %.3fs: %s", elapsed, e)
@@ -643,12 +655,12 @@ class OllamaManager:
 
     async def _chat_stream(self, messages: list, model: str, start: float,
                            temperature: float | None = None, num_gpu: int | None = None,
-                           think: bool = False):
+                           think: bool = False, num_ctx: int | None = None):
         """Async generator streaming tokens, then a final __meta__ stats sentinel."""
         try:
             body: dict = {"model": model, "messages": messages, "stream": True}
             body["think"] = think  # see chat(): reasoning off by default for speed
-            opts = _chat_options(temperature, num_gpu)
+            opts = _chat_options(temperature, num_gpu, num_ctx)
             if opts:
                 body["options"] = opts
             self._apply_keep_alive(body)