feat: openrouter out of the box experience for goose installations #3507
Conversation
michaelneale
added
ui
rust
* main: Extension Library Improvements (#3541) fix(ui): enable selection of zero-config providers in desktop GUI (#3378) refactor: Renames recipe route to recipes to be consistent (#3540) Blog: Orchestrating 6 Subagents to Build a Collaborative API Playground (#3528) Catch json errors a little better (#3437) Rust debug (#3510) refactor: Centralise deeplink encode and decode into server (#3489) feat: deprecate jetbrains extension in favor of public one (#2589) feat: Add LiteLLM provider with automatic prompt caching support (#3380) docs: update desktop instructions for managing sessions (#3522) docs: update desktop instructions for session recipes (#3521) Replace mcp_core::content types with rmcp::model types (#3500) docs: update desktop instructions for tool perms (#3518) docs: update desktop instructions for tool router (#3519) Alexhancock/reapply 3491 (#3515) docs: update mcp install instructions for desktop (#3504) Docs: Access settings in new UI (#3514) feat: switch from mcp_core::Role to rmcp::model::Role (#3488) Revert "fix the output not being visible issue (#3491)" (#3511) fix: Load and Use recipes in new window (#3501)
michaelneale
requested review from
DOsinga,
angiejones,
blackgirlbytes and
zanesq
michaelneale
changed the title
* main: fix: use sequential when sub recipe task is 1. (#3573) fix: track message id to keep like with like (#3572) Replace mcp_core::prompt with rmcp::model types (#3561) feat (ui): close recipe modals with esc key (#3568) feat: recipes can retry with success criteria (#3474) Env var to set Ollama request timeout (#3516) Updating docs to match new UI (#3552) Improve Claude Code provider error message for missing CLI (#3363) feat: Work around Gemini API tool call quirks (#3328) feat(ui): Source CashSans-Bold and improve overall text rendering (#3091) refactor: Use openapi for recipe endpoint types and in frontend (#3548) Fix Google Analytics error for local dev (#3544)
* main: docs: desktop recipe format (#3594) Fix model display name not being updated immediately after leaving settings (#3587) Added option to summarize the chat when an error is triggered (#3598) Remove mcp_macros and unused types (#3581) fix: add fallback id to messages if none provided (#3584) feat: migrate ErrorData from internal mcp crates to rmcp version (#3586) fix: adjust subrecipe description to allow running tests (#3585) Scenario tests (#3430) feat: migrate JsonRpcMessage/Request/Response/Error/Notification from internal mcp crates to rmcp versions (#3564) Restore recipe parameters functionality (#3530) feat: Enhanced loading states with thinking icons and flying bird animation (#3478) Agent loop defensive (#3554) chore: remove needless clone() in goose/providers (#2528) Add recipe install warning (#3537) Replace mcp_core::resource::* with rmcp types (#3563) Add YouTube video embed to using-goosehints.md (#3560) fix: ensure retry-config and success-criteria are populated in openapi spec (#3575)
DOsinga
left a comment
DOsinga
left a comment
There was a problem hiding this comment.
be excited to get this in! I did notice a lot of LLM comments that I think we should start to push back on a bit. similarly the catch all errors approach in ts is something to watch out for
the initial window looks off on my local - text is greyed out. |
|
in the CLI I'm seeing this: 
It doesn't let me pick a model at all |
|
@DOsinga I expect as it is using the lead/worker out of the box, I bet if I simplify it then it would do better at that (although I don't think listing models will work very well with openrouter as there are now so many... not really sure what we do there longer term) |
* main: (27 commits) Add inline python extension (#3107) fix: add maintainer, homepage and categories to DEB/RPM package config (#3096) blog: agent to agent convo (#3677) Possible to disable random thinking messages (#3304) Two VS code tutorials (#3603) small blog fixes (#3549) docs: fix installation command for YouTube Transcript MCP in servers.json (#3595) Docs for using Docker Model Runner as a local LLM provider. (#3509) Docs: VS Code Extension move to tutorials (#3601) Fix working directory when session has no messages (#3513) goose docs MCP server (#3665) Remove confusing status output when testing sharing url connection and it shows 404 (#3659) chore: use typed notifications from rmcp (#3653) feat: convert GetPromptResult from mcp_core to rmcp version (#3650) feat: Replace usage of mcp_core Tools/ToolAnnotations in openapi schema (#3649) fix: ensure execution task result is shown (#3629) docs: Quick spotlight fix (#3633) alexhancock/rmcp-tools-annotations (#3617) fix: clean up subagent (#3565) Adds the `WaitingForUserInput` state (#3620) ...
Looking at how Roo Code works, getting both the full current list and being able to search that list would be ideal for me. If you are doing manual configuration, you probably want the control and choice, I certainly do! The pain was sufficient, that I found myself falling back to gemini cli for now! As a stretch goal which is almost certainly beyond the scope here, I would like support for provider routing, but that's me. At that point, I should probably be using LiteLLM instead to handle my OpenRouter connection. |
* main: blog: streamlining detection development w/ recipes (#3689) fix: have option for cli providers to use their configured or default model (#3683) docs: new blog post and corrections to an old one on goosehints (#3657) Resolve sub recipe path relative to the parent recipe path (#3642) Speed up recipe loading from deeplinks and various fixes (#3662) fix cmd + , not opening settings (#3694) Add warning when JSON env parsing fails. (#3696) chore: refactor session naming into provider (#3678) feat (ui): File picker for scheduling recipes default to recipe dir (#3611) fix: address issue with streamable http interactions via mcp (#3693) Provider scenario tests (#3688) Fix conversations before they hit the LLM (#3660) cli: add detailed instruction for WSL users (#3496) feat: recipe runs will now prompt for missing extension secrets (#3668) fix: pricing integration tests -> trying more runs for cache and retries (#3546)
|
@evangineer yeah - am curious about the routing in openrouter, people can try it, but I think in the end most agents will probably do a better job for now knowing what they need (everyone is trying something though). |
|
@evangineer if you had time - woudl love a issue/write up of a reasonable flow for how we can search/filter the list, yeah it is so long I can't even use it even though it is dynamically loaded in the CLI (but not GUI). We can solve it in the GUI at least I guess... but cli you want to have some TUI that filters it (we can probably also show recommended models). this flow in any case will pick an opinionated default |
|
OK this is finally ready - I think I addressed most things (modulo things that will follow on) - so I think will merge this once it is clean - it really is just for initial experience, not advanced config (and ... what is most painful is I have to blow away my goose config to test and can't run goose while I am mucking around! you realise how much you depend on it!) |
|
tagging @taniashiba @blackgirlbytes as this will be a new "unboxing" flow (convenience) for really new users (which we may want to update notes/docs for when released - shouldn't change anything for existing users). Note that openrouter has ability to top up with things like cashapp - so I wonder if we can make a guide (I couldn't find a URL to open) which helps people create QR code and we can top up their account if starting? 
and then: 
type of thing? |
* main: chore: small refactor on agent.rs (#3703) docs: Add GitMCP Tutorial to Extensions Library (#3716) chore: Speed up CI (#3711) Fix tool vector tests (#3709) docs: GitMCP Tutorial (#3708) Remove unused dependencies (#3626) feat: update Groq models for better tool calling support (#3676) chore: remove ffi libraries and related code (#3699) only run google analytics in prod (#3395) Fix typo in quickstart document (#3447) fix: pricing estimation for OpenRouter in goose-cli (#3675) fix: escape control characters in LLM tool call arguments JSON (#2893) feat(githubcopilot): add ability to fetch supported models (#2717) Create a message ID for tool response messages (#3591) fix: Fixed 404 broken link to extensions page in index.md (#3623)
* main: Blog:How OpenRouter Unlocked Our Workshop Strategy (#3726)
|
this is killing me - doesn't reload correctly when running fully packaged (or this is another regegression that has crept in). whatever it is, will be quite a bit more debugging (with 30 min + roundtrips) ... so not sure yet of the future of this. |
|
ok remaining challenge: the flow will provide config by storing it straight to disk (in core goose crate, which is good practice, but GUI likes to duplicate things so still trying to work out how to make that work). |
* main: Remove unused Memory Mode / Computer-Controller Mode code (#3743) docs: Add Neon MCP Server tutorial (#3639) Update OSX codesigning and notarization (#3658) Fix slow typing in chat input with long running sessions (#3722) Fix html content detection regex to not include markdown autolinks (#3720) Must have missed this one (#3733) Ok, well, that got out of hand (#3718) feat: openrouter out of the box experience for goose installations (#3507)
working on: #3503
this will use PKCE when setting up goose for the first time to automatically setup goose with openrouter.
uses the goose crate for core logic for managing the flow and keys and installing the config, and consumes it from the cli, and also goose-server/desktop use the same logic.
unit tests for the rust side core code.
Please see this short demo of it running the desktop for first time - pops up and logs in to openrouter, then back to goose, ready to go, configured with models and keys, no manual work needed:
openrouter-desktop.mov
This is designed to be the first flow onboard experience for a brand new install of goose.
Upnext:
goose configuretype of flow for when it is brand new so it is same for CLI as desktopTo test:
just run-uiwhich will set it up fresh