add streaming of ai text

2 files changed, 68 insertions, 0 deletions

diff --git a/.rules/changelog/2026-03/24/07.md b/.rules/changelog/2026-03/24/07.md
new file mode 100644
index 0000000..d1327e7
--- /dev/null
+++ b/.rules/changelog/2026-03/24/07.md
@@ -0,0 +1,29 @@
+# Changelog — 2026-03-24 (07)
+
+## Added: Streaming AI Responses
+
+### New
+- **`src/ollama-client.ts`** — Added `sendChatMessageStreaming()` function:
+  - Uses native `fetch()` with `stream: true` for real-time token delivery
+  - `readNdjsonStream()` async generator parses newline-delimited JSON, handles partial lines across chunks
+  - `onChunk` callback fires for every content delta
+  - `AbortSignal` support for cancellation
+  - Full tool-calling agent loop support (tools execute non-streamed, final answer streams)
+  - `StreamingChatOptions` interface for clean parameter passing
+- **Plan file** at `.rules/plan/streaming.md`
+
+### Changed
+- **`src/chat-view.ts`** — Switched from `sendChatMessage` to `sendChatMessageStreaming`:
+  - Creates empty assistant bubble immediately, appends text chunks progressively
+  - `AbortController` created on send, aborted on Stop click or view close
+  - Send button becomes red "Stop" button during streaming
+  - `debouncedScrollToBottom()` limits scroll updates to every 50ms
+  - On completion/abort, streaming class removed from bubble
+- **`styles.css`** — Added `.ai-organizer-stop-btn` red styling for the stop button
+
+### Tool call UI refinements (from previous session)
+- Removed "Tool Use:" prefix from header (icon is sufficient)
+- Removed redundant "Searched for:" / "Read file:" from summaries
+- Added leading `/` to file paths in read_file summary
+- Added result summary line (e.g. "3 results found", "no results found")
+- JSON args and result are now inside a collapsible `<details>` element
diff --git a/.rules/plan/streaming.md b/.rules/plan/streaming.md
new file mode 100644
index 0000000..925a2f1
--- /dev/null
+++ b/.rules/plan/streaming.md
@@ -0,0 +1,39 @@
+# Plan: Streaming AI Responses
+
+## Goal
+Show AI text token-by-token as it arrives, instead of waiting for the full response.
+
+## Key Changes
+
+### 1. `src/ollama-client.ts` — New streaming chat function
+- Add a `sendChatMessageStreaming()` function that sets `stream: true` in the request body.
+- Ollama returns `application/x-ndjson` — each line is a JSON chunk with `message.content` (partial text).
+- The final chunk has `done: true`.
+- Use native `fetch()` API with `response.body.getReader()` + `TextDecoder` to read chunks, split by newlines, parse each JSON line. (`requestUrl()` does not support streaming.)
+- Accept an `onChunk(text: string)` callback that fires for every content delta.
+- Accept an `AbortSignal` for cancellation support.
+- For the tool-calling agent loop: after streaming completes, check the accumulated message for `tool_calls`. If present, execute tools (non-streamed), then stream the next turn. Only the final text turn streams visibly. Tool execution rounds stay non-streamed.
+- Return the full accumulated text at the end.
+
+### 2. `src/chat-view.ts` — Progressive message rendering
+- When sending, create an empty assistant message bubble immediately.
+- Pass an `onChunk` callback that appends each text delta to that bubble's `textContent`.
+- Scroll to bottom on each chunk (debounced to ~50ms to avoid performance issues).
+- While streaming: disable the Send button, show a "Stop" button that can abort the stream via `AbortController`.
+- On stream end or abort: finalize the message, re-enable input.
+
+### 3. `styles.css` — Minor additions
+- Add a blinking cursor indicator on the streaming message bubble (e.g. `::after` pseudo-element).
+- Style the Stop button.
+
+## Considerations
+- **`fetch()` vs `requestUrl()`:** `requestUrl` is Obsidian's abstraction (works on mobile, handles CORS). `fetch` works for localhost calls on desktop. On mobile, `fetch` to `localhost` may not work. Use `fetch` only for streaming and keep `requestUrl` for non-streaming/fallback.
+- **Tool calling + streaming:** Ollama supports streaming with tools. Streamed chunks accumulate `tool_calls` across multiple chunks. Simpler approach: use streaming for the text response; if tool calls come back, fall back to non-streamed agent loop for tool execution, then stream the final response.
+- **Abort handling:** `AbortController` passed to `fetch`. On abort, clean up gracefully — keep partial text visible in the chat but don't add it to message history for future context.
+
+## Estimated file changes
+| File | Change |
+|------|--------|
+| `ollama-client.ts` | Add `sendChatMessageStreaming()`, keep existing `sendChatMessage()` |
+| `chat-view.ts` | New streaming send flow, onChunk handler, stop button, debounced scroll |
+| `styles.css` | Streaming cursor animation, stop button style |