Repository navigation
Release: Lite 0.2.0, OpenCode + Pi, universal warm pool, Open WebUI cleanup - #57
Merged
Merged
Conversation
* Lite 0.2.0: email sign-in, admin-enabled models, no embedded chat - Lite gets LOGIN_EMAIL + PASSWORD (the agent's generated admin login, shown under Access) and AGENT_<ID>_MODELS (the admin-enabled models, named the way each engine's ACP server lists them). LOCK_MODEL only when there is no list. - The dashboard's embedded chat, its /chat-embed route, the embed secret and frame origins are gone; Access shows the chat login instead. - The Controls tab's Model card is gone: the model is picked in the chat. - Reprovision rewrites Lite's whole env, and a box on an older Lite (0.1.0 exits at start) is reinstalled; chat-ui:migrate selects those boxes too. - No AGENT_IDLE_MINUTES (Lite 0.1.0 read it as an agent and exited). Co-Authored-By: Claude <noreply@anthropic.com> * Engines accept the chat UI's models; fail fast on a crash at start - Claude Code: settings.json availableModels; Codex: a model_catalog_json built from the installed Codex's own catalog; Hermes: a pinned, idempotent patch so an openrouter:<slug> switch stays on OpenRouter (0.19.0 moves some slugs to GMI). Each was checked against the pinned engine's ACP server. - The start script no longer retries a service that crashed at start, and the completion log says how long the deploy took and whether it came from the warm pool. Co-Authored-By: Claude <noreply@anthropic.com> * OpenCode and Pi as beta runtimes, driven over ACP - OpenCode (opencode-ai 1.19.2, its own `opencode acp`) and Pi (@earendil-works/pi-coding-agent 1.2.0 through pi-acp 0.0.33), both on OpenRouter, behind CLI_RUNTIMES_ENABLED like Claude Code and Codex. - The CLI gateway gets a generic ACP driver for them (one agent process per turn; resume via session/load; permission requests allowed, as for the SDK engines). Checked live against the real OpenCode and pi-acp. - Neutral placeholder icons until the official marks are added. Co-Authored-By: Claude <noreply@anthropic.com> * Universal warm pool: one image with every chat-UI engine - A universal pool VM bakes Hermes, Claude Code, Codex, OpenCode, Pi, Lite and every ACP side into one bootstrap (30KB, under DigitalOcean's 64KB user-data cap). A deploy of any of them claims it; only the chosen engine's env, config and units are written and started at claim. - It replaces the Hermes-only target (POOL_SIZE_UNIVERSAL, default 2; POOL_SIZE_HERMES still read). The pool key folds in every engine's pin, so a bump on any of them retires the pool. OpenClaw keeps its own pool. - A claim records the baked runtime and Lite versions, so a later reprovision does not reinstall Lite. Co-Authored-By: Claude <noreply@anthropic.com> * Docs: the new engines, Lite's sign-in and models, the universal pool Co-Authored-By: Claude <noreply@anthropic.com> * OpenCode and Pi: offer only models with Hermes' 64k context floor Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com>
A BYOS server prebaked with Open WebUI kept /opt/open-webui (several GB) after an agent was deployed on it; only the service was stopped. Every provision now removes the whole install (unit, /opt/open-webui, env file, user), with the same steps the migration uses. It also stops touching Open WebUI when another agent on the same BYOS server still uses it. Phase B used to stop it on every deploy, so an OpenClaw install there would have killed a not-yet-migrated agent's chat.
This was referenced Oct 23, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Brings
developtomain:AGENT_IDLE_MINUTESas an agent and exited. Pinsagentwebui-lite==0.2.0(on PyPI). Lite sign-in by email + password (Access → Chat Login), the admin-enabled models in the chat (the dashboard's Chat card and Model card are removed), OpenCode and Pi as beta runtimes behindCLI_RUNTIMES_ENABLED, and one universal warm pool image with all five chat-UI engines (POOL_SIZE_UNIVERSAL, default 2)./opt/open-webui, env file, user), unless another agent on the same BYOS server still uses it. It also fixes a Replace Open WebUI with AgentWebUI Lite, embedded in the dashboard #52 bug: any deploy on a shared BYOS server stopped a not-yet-migrated agent's Open WebUI. Rebased onto Lite 0.2.0 sign-in and models, OpenCode + Pi, universal warm pool #56. The conflicts were the embed-secret code that Lite 0.2.0 sign-in and models, OpenCode + Pi, universal warm pool #56 removed.Schema: unchanged. No dependency,
.env.exampleordeploy/changes.Before deploying: set
KEEP_FAILED_VMS=false(or remove it) in the prod.env.After deploying:
/admin/instances→ Pre-Baked VMs. Old Hermes-only pool VMs are retired, and universal VMs bake.npm run chat-ui:migrate(dry run), thennpm run chat-ui:migrate -- --apply. This moves Open WebUI agents and repairs boxes stuck on Lite 0.1.0./admin/models, then deploy one agent per engine. The log should readDeployment complete in Ns (warm pool). Sign in at the agent URL, switch models and send a message.Not verified: no real agent server has run #56 or #55 yet.
🤖 Generated with Claude Code