main
md 31 lines 1.42 KB
Rendered Raw
1 # Message Loop Result Extensions DOX
2
3 ## Purpose
4
5 - Own normalization and policy handling after a model turn completes, before default assistant-history and tool-dispatch processing.
6
7 ## Ownership
8
9 - Extensions receive mutable `result_data` with `llm_result` and may set `skip_default_processing` after fully handling the turn.
10 - `_20_empty_response.py` retries turns with neither response nor reasoning, counts them toward the unusable-response limit without adding a warning to model history, and uses `fw.msg_empty_response.md` for agent-prefixed UI warning text only.
11 - `_30_repeat_response.py` retries response content that exactly matches `loop_data.last_response`, regardless of reasoning, using `fw.msg_repeat.md` for history and `fw.msg_repeat_response.md` for the agent-prefixed UI warning text.
12
13 ## Local Contracts
14
15 - Files run in deterministic filename order.
16 - A handler that sets `skip_default_processing` owns needed history and UI side effects for that turn.
17 - Handlers that should not add side effects after an earlier extension has handled the result must return when `skip_default_processing` is set.
18 - Do not use this point to mutate streamed partial content.
19
20 ## Work Guidance
21
22 - Normalize a completed result before policy extensions compare or persist it.
23 - Keep loop-control policy independent from optional plugins.
24
25 ## Verification
26
27 - Run message-loop and unusable-response regression tests.
28
29 ## Child DOX Index
30
31 No child DOX files.