feat(mimo): native adapter takes the bare mimo name (single mimo)#47
Merged
Conversation
46b2ef9 to
9689481
Compare
added 2 commits
July 4, 2026 02:23
`mimo` existed twice: benchflow core's `mimo-proxy acp` wrapper shim and this
package/manifest ("mimo-acp") running upstream's NATIVE `mimo acp` ACP server
directly. Consolidate on the upstream-native adapter per the naming law (the
ACP path owns bare names): the manifest now declares name "mimo" (dirname
mimo-acp intentionally differs — the loader keys on the declared name), and
"mimo-acp"/"mimo-code" stay as aliases forever. Via benchflow's decoupled-
manifest mechanism (merge_shim_only), deploying this manifest makes the native
definition THE `mimo`, overriding the core wrapper's data fields; core remains
only a fallback when the manifest is absent.
Companion: benchflow adds a transient-5xx retry policy at the LiteLLM proxy —
the known mimo flake (upstream mimo silently swallows a provider 500 into a
0-event turn) never reaches the agent once the proxy absorbs transient 500s.
…tive) The count was FOUR mimo definitions across the ACP path: benchflow core's mimo-proxy wrapper entry, the repo's acp/mimo wrapper-flavor decoupled manifest (agents#45 sync), the acp/mimo-acp native-flavor manifest, and the mimo_acp python package (entry-point flavor of the same native adapter). The ACP-registry model (agentclientprotocol.com) is one declarative entry per agent. Consolidate to exactly one repo definition: acp/mimo/manifest.toml now carries the NATIVE `mimo acp` adapter (name `mimo`, aliases `mimo-acp`/`mimo-code`, dirname == agent name per repo convention). Deleted: the acp/mimo-acp directory (manifest + python package + tests + its CI workflow/lint step). The contract duplicate-name test — which caught the collision — passes; only the pre-existing corust-agent manifest failure remains (present on main). benchflow core's wrapper entry remains as the out-of-repo fallback the decoupled-manifest mechanism overrides; removing it is a benchflow-side follow-up (entangled with 6 test files + integration CI configs).
9689481 to
4a5271d
Compare
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.
Consolidates mimo onto one declarative entry — the ACP-registry model: one agent, one manifest, generic runtime.
Before: FOUR mimo definitions in the ACP path
mimo—mimo-proxy acpwrapper entry (registry.py)acp/mimo/manifest.toml— wrapper-flavor decoupled manifest (agents#45 sync of 1)acp/mimo-acp/manifest.toml— nativemimo acpmanifestmimo_acppython package — entry-point flavor of 3After: ONE
acp/mimo/manifest.toml= the native adapter (namemimo, aliasesmimo-acp/mimo-codeforever; dirname == agent name per repo convention). Pure data — loads viaBENCHFLOW_AGENTS_DIRor manifest auto-load, no install.acp/mimo-acp/(manifest + python package + tests), its CI workflow + lint step.test_no_duplicate_agent_names— which caught the 2-manifests-one-name collision — passes; the only remaining contract failure is the pre-existingcorust-agentTOML issue (present on main).Verified working (daytona, citation-check, with-skill)
Native manifest + benchflow#882's transient-5xx retry (the make-it-work fix for the documented silent-swallow flake): reward 1.0 (acp 51 / llm 13, no error) — the previously-flaky native adapter now scores clean.
Companion: benchflow#882.