Skip to content

feat: /model slash command with native Discord autocomplete#186

Open
sh14y wants to merge 5 commits intoopenabdev:mainfrom
sh14y:feat/model-switch-command
Open

feat: /model slash command with native Discord autocomplete#186
sh14y wants to merge 5 commits intoopenabdev:mainfrom
sh14y:feat/model-switch-command

Conversation

@sh14y
Copy link
Copy Markdown

@sh14y sh14y commented Apr 10, 2026 •
edited
Loading

Design discussion: https://discord.com/channels/1491295327620169908/1491365157010542652/1493920430283686049

Summary

Adds runtime model switching via Discord's native /model slash command, with autocomplete-driven model selection.

What changed

This PR contains three commits that build on each other:

  1. feat: add !model command for runtime model switching via session/set_model

    • Wires up the ACP session/set_model JSON-RPC method (verified on Kiro CLI v1.29.3)
    • Parses models.availableModels from session/new response
    • Adds alias resolution: auto, opus, sonnet, haiku
    • Initial text-based !model interface (superseded in commit 3 below)

  2. fix: self-bootstrap kiro creds and config.toml inside the binary

    • Required for the cloud deployment we're using to test this; restores credentials from KIRO_CRED_B64 and generates config.toml from env vars when missing. Idempotent — no-op when files already exist.

  3. feat: replace !model text command with /model slash command + autocomplete ← the main UX change

    • Removes the !model text intercept entirely
    • Registers /model as a guild command on ready() for instant propagation
    • Adds interaction_create handler for both Command and Autocomplete interactions
    • SessionPool gains a cached_models snapshot updated on every session/new, so autocomplete can answer within Discord's 3-second deadline without spawning a fresh agent (~10s cold start)
    • Background warmup task at startup populates the model cache before first use
    • Aliases (auto, opus, sonnet, haiku) appear in autocomplete as friendly shortcuts only when their target model is actually available
    • The set path defers the response since cold session creation can exceed 3s

UX

Action Result
Type /m in Discord Native command picker shows /model
Tab into the model field Autocomplete returns aliases + all available model ids (current model marked)
Pick with ↑↓, hit enter Bot defers, sets the model on the channel/thread session, replies ✅ Switched to <id>
/model with no argument Bot replies with the full list and marks the current selection

Important: bot invite scope

/model will not appear unless the bot is invited with both bot and applications.commands OAuth scopes. Existing installations with only bot scope must re-invite using a new URL — the bot account stays the same and no data is lost.

The README has been updated to call this out in the Quick Start section.

Test plan

  • cargo build --release clean
  • /model (no arg) lists all available models with current marker — verified
  • /model auto switches and replies ✅ Switched to auto — verified
  • Verify /model sonnet resolves alias and switches
  • Verify autocomplete shows live filtered suggestions as you type
  • Allowlist (channel + user) enforcement for the slash command + autocomplete path — code-level via pure allowlist_decision() + 9 unit tests (empty lists, channel in/out, thread parent in/out, user in/out, combinations). Runtime smoke test deferred to post-deploy.

@sh14y
feat: add !model command for runtime model switching via session/set_…
caac7a3 
…model

- Add ModelInfo struct and model state fields to AcpConnection
- Parse available models from session/new response
- Add session_set_model() and resolve_model_alias() methods
- Add !model Discord command to list and switch models
- Support aliases: opus, sonnet, haiku, auto

Kiro ACP session/set_model verified on Kiro CLI v1.29.3
@sh14y sh14y requested a review from thepagent as a code owner April 10, 2026 14:29
@sh14y sh14y force-pushed the feat/model-switch-command branch from 5494848 to caac7a3 Compare April 10, 2026 15:28
@sh14y
fix: self-bootstrap kiro creds and config.toml inside the binary
9151206 
Move the entrypoint.sh logic into main.rs so openab is independent of any
external shell wrapper. Before loading config, it now:

  1. Restores ~/.local/share/kiro-cli/data.sqlite3 from KIRO_CRED_B64 if
     missing (idempotent — skips when file already exists, e.g. from a
     mounted volume).
  2. Generates the config file from DISCORD_BOT_TOKEN / DISCORD_CHANNEL_ID
     env vars when the path doesn't exist. Secrets are written as literal
     ${VAR} placeholders so they get expanded by load_config() at read
     time and never land on disk in plaintext.

Why: Zeabur deployments were intermittently running images built without
the entrypoint shim (e.g. when the source branch was switched mid-flight),
producing "failed to read /etc/openab/config.toml" with no entrypoint logs.
With the bootstrap inside the binary, the deploy is robust to any image
build path that happens to omit our shell wrapper.

The Dockerfile entrypoint.sh remains as a belt-and-suspenders no-op when
present.
@sh14y sh14y force-pushed the feat/model-switch-command branch from a4290bb to 9151206 Compare April 11, 2026 01:51
@sh14y
feat: replace !model text command with /model slash command + autocom…
bdc886d 
…plete

Discord's slash commands give a much better UX than text commands:
- Typing /m surfaces /model in Discord's native command picker
- Autocomplete on the model option shows the live list of available
  models (and aliases) with arrow-key selection
- Discord renders the response as an interaction reply instead of
  cluttering the channel as a regular bot message

Implementation notes:

- SessionPool gains a cached_models snapshot updated on every
  session_new(). Autocomplete must respond within Discord's 3-second
  deadline, which rules out spawning a fresh kiro-cli (~10s cold start).
- main.rs adds a background warmup task that creates one session at
  startup so the cache is populated before the first user invocation.
- /model is registered as a guild command on ready() for instant
  propagation (global commands take up to 1 hour).
- The interaction handler defers the response on the set path because
  cold session creation can exceed Discord's 3s window for non-deferred
  replies.
- README documents the applications.commands scope requirement; without
  it the slash command will not appear in Discord clients.

The !model text command is removed entirely — /model is its replacement,
not an addition.
@sh14y sh14y mentioned this pull request Apr 11, 2026
Closed
7 tasks
@sh14y sh14y changed the title feat: add !model command for runtime model switching via session/set_model feat: /model slash command with native Discord autocomplete Apr 11, 2026
This was referenced Apr 11, 2026
Closed
Closed
@thepagent
Copy link
Copy Markdown
Collaborator

thepagent commented Apr 15, 2026

Native Discord slash command with autocomplete and model caching is a nice feature but touches multiple components (Discord API registration, SessionPool, autocomplete). Could you please create a Discord discussion thread and add the link to this PR description for design tracking? Thanks! 🙏

@thepagent
Copy link
Copy Markdown
Collaborator

thepagent commented Apr 15, 2026

DC https://discord.com/channels/1491295327620169908/1491365157010542652/1493920430283686049

@thepagent
Copy link
Copy Markdown
Collaborator

thepagent commented Apr 15, 2026

Related issues worth noting alongside this PR:

The cached_models approach for beating Discord's 3-second autocomplete deadline is a clean solution. One thing still open per the test plan: allowlist enforcement (channel + user) for the slash command path — worth tracking before merge.

@sh14y @claude
fix: enforce channel+user allowlist on /model slash command and autoc…
90a5dd5 
…omplete

- Add interaction_allowlist_ok() helper (channel, thread-parent, user checks)
- Apply to handle_model_command: reject with ephemeral reply if not allowed
- Apply to handle_model_autocomplete: return empty choices if not allowed

Addresses review feedback from thepagent on PR openabdev#186

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@sh14y
Copy link
Copy Markdown
Author

sh14y commented Apr 15, 2026

Thanks for the review! I've addressed both points:

@sh14y @claude
test: add unit tests for allowlist_decision
583f6d3 
Extract pure allowlist logic into ctx-free allowlist_decision() so it can be
unit-tested without a Discord runtime. interaction_allowlist_ok() becomes a
thin wrapper that resolves the thread parent via HTTP and delegates.

Covers: empty lists, channel in/out, thread parent in/out, user in/out, and
the "thread-allowed + user-denied" combination.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@sh14y
Copy link
Copy Markdown
Author

sh14y commented Apr 15, 2026

Follow-up: extracted the allowlist logic into a pure allowlist_decision() function and added 9 unit tests covering:

  • Empty allowlists allow everything
  • Channel in/out of allowlist
  • Thread parent in/out of allowlist
  • User in/out of allowlist
  • Combinations (e.g. thread-allowed + user-denied)

interaction_allowlist_ok() is now a thin wrapper that resolves the thread parent via HTTP and delegates to the pure function. All 9 tests pass via cargo test. Runtime verification against a live Discord instance is deferred until post-deploy — the logic is identical to the existing message() path which has been running in prod.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@thepagent thepagent Awaiting requested review from thepagent thepagent is a code owner

At least 2 approving reviews are required to merge this pull request.

Assignees

No one assigned

Labels

pending-maintainer

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@sh14y @thepagent @chaodu-agent