feat(proxy): Bob (BobShell) gateway — reduce model calls, pass control-plane through#12
Merged
Conversation
added 3 commits
July 21, 2026 14:45
Bob is OpenAI-compatible but calls Bob-specific paths. Add an opt-in gateway (enabled by BOB_UPSTREAM / --bob-upstream): - POST /inference/v1/chat/completions is reduced through the pipeline like any OpenAI chat and forwarded to the Bob backend, passing Bob's own auth through. - A "/" catch-all transparently proxies Bob's control-plane calls (/admin/v1/profile, /inference/v1/model/info, …) verbatim so the CLI boots and authenticates. Less specific than every existing route, and only registered when BOB_UPSTREAM is set, so default behavior is unchanged. Point Bob's CUSTOM_BASE_URL at this proxy. Verified end-to-end against the Bob backend with deterministic components. Ports winnow's bob profile. Assisted-By: Claude <noreply@anthropic.com> Signed-off-by: Osher-Elhadad <Osher.Elhadad@ibm.com>
The kagenti->rossoctl rename string-replaced import paths without re-sorting, so github.com/rossoctl/... now sorts after github.com/maximhq/...; gofmt re-orders the import groups. Fixes `make lint` (gofmt check) on this branch. Assisted-By: Claude <noreply@anthropic.com> Signed-off-by: Osher-Elhadad <Osher.Elhadad@ibm.com>
Add an integrations section covering the opt-in Bob gateway: enable with BOB_UPSTREAM, point Bob's CUSTOM_BASE_URL at the proxy, model calls reduced / control-plane passed through. Notes the verified deterministic setup. Assisted-By: Claude <noreply@anthropic.com> Signed-off-by: Osher-Elhadad <Osher.Elhadad@ibm.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Adds an opt-in Bob (BobShell) gateway to the proxy so IBM Bob's traffic can be routed through context-guru — the way the winnow prototype already does. Enabled only when
BOB_UPSTREAM(or--bob-upstream) is set; default proxy behavior is unchanged.Why
Bob is OpenAI-compatible via
CUSTOM_BASE_URL, but it calls Bob-specific paths: the model call isPOST /inference/v1/chat/completions, and it makes control-plane calls (GET /admin/v1/profile,/inference/v1/model/info, …) that must reach the backend unmodified or the CLI won't boot. The proxy previously exposed only/openai/v1/chat/completionsand/anthropic/v1/messages, so Bob couldn't be pointed at it.How it works
POST /inference/v1/chat/completions→ reduced through the pipeline like any OpenAI chat, forwarded to the same path onBOB_UPSTREAM; Bob's own auth passes through (no key injection)./catch-all → transparently proxies Bob's control-plane calls verbatim toBOB_UPSTREAM. Less specific than every existing route, and only registered whenBOB_UPSTREAMis set.Usage
Verified
proxytest: model call on/inference/v1/chat/completionsis reduced (dedup) and forwarded to the right path; aGET /admin/v1/profilecontrol-plane call is passed through verbatim.format+toon): Bob authenticated and answered correctly through the proxy, with its model call reduced.Notes
chore: gofmtcommit fixing repo-wide unsorted imports left by the kagenti→rossoctl rename (same fix as docs: Material for MkDocs site with GitHub Pages deploy #11); it coalesces on merge.docs/integrations.md.