fix(mcp): sanitize for gemini when model is unresolved (#697)

* fix(mcp): sanitize for gemini when model is unresolved

isGeminiRouted() previously required the effective model string to
contain "gemini" — but when payload.model="auto" (or any unresolved
slug) reaches addTools(), `effective` is the literal "auto", which
doesn't match. opencode then auto-selects gemini *after* the MCP
server has registered raw arktype schemas, and every tool turn dies
on `function_declarations[*].properties[*].any_of[*].enum: only
allowed for STRING type`.

widen the gate: any unresolved specifier (undefined / "auto" / a
slug without a `provider/` prefix) is treated as gemini-routed and
sanitized. the transforms are universally compatible normalizations
so the false-positive cost is negligible. tighten case 3 to preserve
`description` so the only lossy path no longer drops operator-facing
context.

fixes #676.

* revert case-3 description preservation

per pullfrog review on #697: keeping `description` as a peer of
`anyOf`/`oneOf` directly contradicts the file's own header (lines
19-21) and the upstream opencode #14659 rationale that gates this
sanitizer — gemini requires anyOf to be the ONLY field on a schema
node, sibling keywords trigger
`anyOf must be the only field in a schema node`. the change was
speculative scope creep with no evidence, and would silently
re-introduce a different gemini failure for any future schema using
`.describe().or(...)`. the bug fix for #676 doesn't need it (arktype
doesn't emit non-collapsible anyOf for current tool schemas).
This commit is contained in:
Colin McDonnell
2026-05-13 02:31:59 +00:00
committed by pullfrog[bot]
parent 43bb14bf87
commit 159389fad2
+25 -6
View File
@@ -176,13 +176,32 @@ export function sanitizeToolForGemini<T extends Tool<any, any>>(tool: T): T {
}
/**
* true when the effective upstream model is served by google's generative
* language API — directly (`google/*`), via opencode (`opencode/gemini-*`),
* or via openrouter (`openrouter/google/gemini-*`). slug-substring match
* works because every gemini route's model id contains "gemini".
* true when the effective upstream model is — or might become — google
* generative language API traffic. matches:
* - direct `google/*`, opencode `opencode/gemini-*`, openrouter
* `openrouter/google/gemini-*` (slug substring "gemini" wins).
* - any unresolved specifier: `undefined`, `"auto"`, or a slug that
* didn't map through the alias registry (no `provider/` prefix).
* these flow through the agent's own auto-select, which may land
* on gemini *after* the MCP server has already registered tools —
* at which point sanitization is too late to apply. erring on the
* side of sanitizing is safe: cases 1 + 2 are universally
* compatible JSON-Schema normalizations (enum-only → typed string,
* collapsible const-unions → string enum); case 3 is gemini-
* specific but only fires on non-collapsible unions, which arktype
* does not emit for our current tool schemas. see issue #676 for
* the prod failure that motivated this widening.
*/
export function isGeminiRouted(ctx: ToolContext): boolean {
const effective = ctx.payload.proxyModel ?? ctx.resolvedModel ?? ctx.payload.model;
if (!effective) return false;
return effective.toLowerCase().includes("gemini");
if (!effective) return true;
const normalized = effective.toLowerCase();
if (normalized.includes("gemini")) return true;
// every concrete model resolved through the registry carries a
// `provider/` prefix (e.g. "anthropic/claude-opus-4-7"). anything
// without a slash is either the literal `"auto"` alias or an
// unrecognized slug that resolveModel logged a warning for — both
// route through the agent's late auto-select, which may pick gemini.
if (!normalized.includes("/")) return true;
return false;
}