fix(auto-reply): tighten block streaming defaults
This commit is contained in:
@@ -57,11 +57,11 @@ Block streaming sends partial replies as the model produces text blocks.
|
||||
Chunking respects provider text limits and avoids splitting fenced code.
|
||||
|
||||
Key settings:
|
||||
- `agents.defaults.blockStreamingDefault` (`on|off`)
|
||||
- `agents.defaults.blockStreamingDefault` (`on|off`, default off)
|
||||
- `agents.defaults.blockStreamingBreak` (`text_end|message_end`)
|
||||
- `agents.defaults.blockStreamingChunk` (`minChars|maxChars|breakPreference`)
|
||||
- `agents.defaults.blockStreamingCoalesce` (idle-based batching)
|
||||
- Provider overrides: `*.blockStreaming` and `*.blockStreamingCoalesce`
|
||||
- Provider overrides: `*.blockStreaming` and `*.blockStreamingCoalesce` (non-Telegram providers require explicit `*.blockStreaming: true`)
|
||||
|
||||
Details: [Streaming + chunking](/concepts/streaming).
|
||||
|
||||
|
||||
Reference in New Issue
Block a user