fix: make schema normalization provider-aware for Gemini compatibility

Two bugs in convertTools() caused Gemini's OpenAI-compatible endpoint
to reject tool schemas with 400 "schema requires unspecified property":

1. The Agent tool patch unconditionally pushed 'message' into required[]
   even though 'message' is not a property of the Agent schema. Gemini
   strictly validates that every key in required[] exists in properties.

2. normalizeSchemaForOpenAI() added all property keys to required[] for
   OpenAI strict mode, but this conflicts with Gemini's stricter schema
   validation which rejects required keys absent from properties.

Fix:
- Agent tool patch now only adds a key to required[] if it exists in
  schema.properties (fixes the 'message' 400 error on Gemini)
- normalizeSchemaForOpenAI() accepts a strict flag: true for OpenAI
  (promotes all property keys into required[]), false for Gemini
  (filters required[] to only keys present in properties)
- convertTools() detects CLAUDE_CODE_USE_GEMINI and passes strict=false

Fixes #82

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/services/api/openaiShim.ts

@@ -231,29 +231,47 @@ function convertMessages(
  * which causes 400 errors on OpenAI/Codex endpoints. This normalizes the
  * schema by ensuring `required` is a superset of `properties` keys.
  */
-function normalizeSchemaForOpenAI(schema: Record<string, unknown>): Record<string, unknown> {
+function normalizeSchemaForOpenAI(
+  schema: Record<string, unknown>,
+  strict = true,
+): Record<string, unknown> {
   if (schema.type !== 'object' || !schema.properties) return schema
   const properties = schema.properties as Record<string, unknown>
   const existingRequired = Array.isArray(schema.required) ? schema.required as string[] : []
+  // OpenAI strict mode requires every property to be listed in required[].
+  // Gemini rejects schemas where required[] contains keys absent from properties,
+  // so only promote keys that actually exist in properties.
+  if (strict) {
     const allKeys = Object.keys(properties)
     const required = Array.from(new Set([...existingRequired, ...allKeys]))
     return { ...schema, required }
+  }
+  // For Gemini: keep only existing required keys that are present in properties
+  const required = existingRequired.filter(k => k in properties)
+  return { ...schema, required }
 }
 
 function convertTools(
   tools: Array<{ name: string; description?: string; input_schema?: Record<string, unknown> }>,
 ): OpenAITool[] {
+  const isGemini =
+    process.env.CLAUDE_CODE_USE_GEMINI === '1' ||
+    process.env.CLAUDE_CODE_USE_GEMINI === 'true'
+
   return tools
     .filter(t => t.name !== 'ToolSearchTool') // Not relevant for OpenAI
     .map(t => {
-    // Estraiamo lo schema
-    const schema = (t.input_schema ?? { type: 'object', properties: {} }) as any;
+      const schema = { ...(t.input_schema ?? { type: 'object', properties: {} }) } as Record<string, unknown>
 
-    // PATCH PER CODEX: Se è lo strumento Agent, forziamo i campi obbligatori
+      // For Codex/OpenAI: promote known Agent sub-fields into required[] only if
+      // they actually exist in properties (Gemini rejects required keys absent from properties).
       if (t.name === 'Agent' && schema.properties) {
-      if (!schema.required) schema.required = [];
-      if (!schema.required.includes('message')) schema.required.push('message');
-      if (!schema.required.includes('subagent_type')) schema.required.push('subagent_type');
+        const props = schema.properties as Record<string, unknown>
+        if (!Array.isArray(schema.required)) schema.required = []
+        const req = schema.required as string[]
+        for (const key of ['message', 'subagent_type']) {
+          if (key in props && !req.includes(key)) req.push(key)
+        }
       }
 
       return {
@@ -261,7 +279,7 @@ function convertTools(
         function: {
           name: t.name,
           description: t.description ?? '',
-        parameters: normalizeSchemaForOpenAI(schema),
+          parameters: normalizeSchemaForOpenAI(schema, !isGemini),
         },
       }
     })