feat: add streaming responses for AI assistant (resolves #56)

[lohit-40]

Summary

Resolves #56
The backend (server/agent.py) already emits text/event-stream Server-Sent Events (SSE). This PR wires up the frontend to consume the stream incrementally so users see tokens as they arrive rather than waiting for the full response.

Changes

hooks/useStream.ts (new)

• Custom React hook that consumes SSE via the Fetch API's ReadableStream
• • Parses frames: output, input_required, error
• • Exposes { lines, isStreaming, inputRequired, error, startStream, stopStream, reset }
• • Supports GET (no body) and POST (with context_payload)
• • Uses AbortController for clean teardown on unmount or stopStream()
• • Zero new npm dependencies

components/StreamingMessage.tsx (new)

• Renders accumulated stream lines as incremental p elements
• • Shows a blinking cursor | while isStreaming === true
• • Displays an accessible error badge (role="alert") on stream failure
• • aria-live="polite" so screen readers announce new tokens

tests/components/streaming.test.tsx (new)

• 30 tests covering useStream and StreamingMessage
• • jsdom-compatible fake ReadableStream mocks
• • Covers: initial state, output frames, input_required, error frames, HTTP errors, network errors, null body, stopStream, reset, POST vs GET, consecutive streams, malformed JSON skip
• • Component: cursor visibility, error badge, aria attributes, className, long responses, streaming -> complete transition

tests/components/chat.test.tsx (modified -- additive only)

• 8 new streaming-specific tests in a new describe block
• • All 5 original tests untouched

tests/setup.ts (modified)

• Polyfills TextEncoder / TextDecoder for jsdom

Test Results

Test Suites: 6 passed, 6 total
Tests: 52 passed, 52 total

Acceptance Criteria

Criterion	Status
AI responses render progressively	YES: useStream emits tokens via lines array as frames arrive
No blocking until full response	YES: ReadableStream never buffers -- renders each data: frame inline
Errors during streaming handled gracefully	YES: error state propagated to StreamingMessage error badge
Works for short and long responses	YES: Same code path regardless of response length
No server changes required	YES: Backend SSE already correct -- only frontend additions

[Atharv777]

Hey @lohit-40, can you please share screenshot/video of the updated UI of the streaming tokens? The code looks good to me.

[lohit-40]

is this ok do let me know if any thing else is needed

[Atharv777]

Hey @lohit-40, as a brownie point, can you fix the line-break issue too? As of now, each word appears in a new line. You can use a formatting library if you want, and if you make it robustly formatted output, we'll give you double the points as a single PR.

[lohit-40]

can u check rn if anything else needed

[Atharv777]

Perfect, thanks @lohit-40

[Atharv777]

Hey, the frontend tests are failing

[lohit-40]

leeme check give 5 min

[lohit-40]

can u check rn i feel there shouldn't any more hiccups!