docs: update agent config reference, custom provider api_type, and slash command behavior#2999
Conversation
Adds `use_commands` and `use_skills` to the agent properties reference and full schema YAML example in docs/configuration/agents/index.md. These agent-level fields were introduced by #2971 but were missing from the agent reference doc. Also fixes a broken example link in docs/configuration/overview/index.md that pointed to the wrong repo (cagent instead of docker-agent). Source PR: #2971
Updates docs/providers/custom/index.md to reflect that `api_type` is now auto-detected based on the model name for custom OpenAI-compatible providers. Newer models (gpt-4.1, o-series, gpt-5, Codex) default to `openai_responses`; others default to `openai_chatcompletions`. Previously the docs showed a hardcoded default of `openai_chatcompletions` which was incorrect post-merge. Source PR: #2985
Documents the fix from #2979: named slash commands now bypass the message queue and execute immediately, even while the agent is busy. Updated docs/configuration/agents/index.md (Named Commands section) and docs/features/tui/index.md (slash command usage) to clarify the distinction between immediate commands and queued chat messages. Source PR: #2979
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
Documentation changes are accurate and internally consistent. The broken link fix ( → ) is correct, use_commands/use_skills are properly documented in both the YAML schema example and the properties table, and the api_type auto-detection is clearly described in docs/providers/custom/index.md. One minor wording inconsistency was found between the two slash command behavior notes (see inline comment).
|
|
||
| Define reusable prompt shortcuts that can send prompts to the current agent or switch to a different sub-agent: | ||
|
|
||
| > **Note:** Named slash commands execute immediately, even while the agent is processing another message. Unlike regular chat messages (which are queued), slash commands interrupt the current operation and run right away. |
There was a problem hiding this comment.
[LOW] Minor wording inconsistency in slash command behavior notes
Two + lines added in this PR describe the same feature with slightly different language:
docs/configuration/agents/index.md(line 267): "slash commands interrupt the current operation and run right away"docs/features/tui/index.md(added): "you can invoke a slash command to interrupt or direct the agent even while it is mid-response"
The agents doc implies the current operation is always interrupted/stopped, while the TUI doc is more nuanced ("interrupt or direct"). Readers consulting both pages may be confused about whether the current operation is always cancelled or can be redirected without cancellation.
Consider aligning the phrasing. For example, using "interrupt or direct the agent" in both places would be more consistent.
Resolves minor inconsistency flagged in PR review: both notes now use 'interrupt or direct the agent' rather than 'interrupt the current operation', which is more accurate and consistent. Source PR: #2979
Documentation updates for recently merged features
This PR brings the
/docscontent in sync with three features merged intomain.Changes
use_commands/use_skillsto agent reference; fix broken example linkapi_typeauto-detection for custom OpenAI providersDetails
docs/configuration/agents/index.md— Addsuse_commandsanduse_skillsto the Full Schema YAML example and Properties Reference table (from #2971). Also adds a note that named slash commands execute immediately (from #2979).docs/configuration/overview/index.md— Fixes broken link pointing to wrong repo (cagent→docker-agent) in the shared skills/commands example (from #2971).docs/providers/custom/index.md— Updatesapi_typedefault from hardcodedopenai_chatcompletionstoauto (model-dependent)and documents the auto-detection logic for newer OpenAI models (from #2985).docs/features/tui/index.md— Adds note about immediate slash command execution vs queued chat messages (from #2979).