v1.7.0

DeepDoc 1.7.0 adds real token-by-token streaming to the chatbot's Fast and Deep Research modes, delivering the same live-answer experience already available in Code-aware mode.

Added

• Added POST /query/stream SSE endpoint that streams the Fast mode answer token-by-token before emitting a final result event.
• Added POST /deep-research/stream SSE endpoint that streams the Deep Research synthesis answer token-by-token before emitting a final result event.
• Added complete_stream() method to LiteLLMChatClient using litellm.completion(stream=True), yielding token strings as they arrive.
• Added token_callback parameter to _complete_with_continuation(), query(), deep_research(), and _run_research_mode() so the final answer generation can push tokens to any caller.
• Added synthesis_token_callback to DeepResearcher so only the synthesis step streams tokens (sub-question expansions remain non-streaming).

Changed

• Updated the generated chatbot UI so Fast and Deep modes fetch from the new /stream endpoints and progressively render the answer with ReactMarkdown as tokens arrive, falling back to the non-streaming endpoints if the stream is unavailable.