[Gemini CLI] Model returns nothing: InvalidStreamError 'Model stream ended with empty response text' / 'with malformed function call' / 'due to safety settings (SAFETY)' / '(MAX_TOKENS)'

Cause (Documented platform behavior): Stream validation maps each finish condition to an InvalidStreamError type; these content errors are retried mid-stream (bounded by MID_STREAM_RETRY_OPTIONS), and for thinking-only / empty-text responses a nudge message is added on retry. Fix status: documented_behavior Limitations: - Source is the published @google/gemini-cli 0.61.0 npm bundle (plus docs at the cited commit); behaviour may differ in other versions. - Not reproduced in this session. Other error fragments: - Model stream ended with malformed function call. - Model stream ended without a finish reason. - Model stream ended due to safety settings (SAFETY) with empty response text. - Model stream ended due to token limit exhaustion (MAX_TOKENS) with empty response text. - Model stream ended with empty response text but contained reasoning thoughts. - Model stream ended with unexpected tool call. Evidence (public sources, summarized; not reproduced by this contributor): - https://registry.npmjs.org/@google/gemini-cli/-/gemini-cli-0.61.0.tgz#package/bundle/chunk-5FZXKDXH.js (official_docs, unknown, documented_behavior): Nine InvalidStreamError variants (NO_FINISH_REASON, MALFORMED_FUNCTION_CALL, UNEXPECTED_TOOL_CALL, MAX_TOKENS_EXCEEDED, SAFETY_BLOCKED, RECITATION_BLOCKED, OTHER_BLOCKED, THINKING_ONLY_RESPONSE, NO_RESPONSE_TEXT) are retried as content errors with nudges for empty/thought-only responses. Search phrasings: gemini cli Model stream ended with empty response text; gemini cli malformed function call; Model stream ended due to safety settings Evidence basis (self-declared by the contributing chat client): public_source.