refactor(vscode): drop unused supportsFim / generateResponse from AutocompleteModel

supportsFim() unconditionally returned true and generateResponse() only
ever threw — leftovers from when the model interface tried to abstract
over FIM-capable vs chat-only providers. Both supported autocomplete
models (Codestral, Mercury) expose FIM, so neither abstraction is needed.

packages/kilo-vscode/src/services/autocomplete/AutocompleteModel.ts

@@ -4,18 +4,6 @@ import type { KiloConnectionService } from "../cli-backend"
 const DEFAULT_MODEL = "mistralai/codestral-2508"
 const PROVIDER_DISPLAY_NAME = "Kilo Gateway"
 
-/** Chunk from an LLM streaming response */
-export type ApiStreamChunk =
-  | { type: "text"; text: string }
-  | {
-      type: "usage"
-      totalCost?: number
-      inputTokens?: number
-      outputTokens?: number
-      cacheReadTokens?: number
-      cacheWriteTokens?: number
-    }
-
 export class AutocompleteModel {
   private connectionService: KiloConnectionService | null = null
   public profileName: string | null = null
@@ -34,10 +22,6 @@ export class AutocompleteModel {
     this.connectionService = service
   }
 
-  public supportsFim(): boolean {
-    return true
-  }
-
   /**
    * Generate a FIM (Fill-in-the-Middle) completion via the CLI backend.
    * Uses the SDK's kilo.fim() SSE endpoint which handles auth and streaming.
@@ -102,20 +86,6 @@ export class AutocompleteModel {
     }
   }
 
-  /**
-   * Generate response via chat completions (holefiller fallback).
-   * Not used when FIM is supported, but kept for compatibility.
-   */
-  public async generateResponse(
-    systemPrompt: string,
-    userPrompt: string,
-    onChunk: (chunk: ApiStreamChunk) => void,
-  ): Promise<ResponseMetaData> {
-    // FIM is the primary strategy; this method is a fallback.
-    // For now, throw — callers should use generateFimResponse via supportsFim().
-    throw new Error("Chat-based completions are not supported via CLI backend. Use FIM (supportsFim() returns true).")
-  }
-
   public getModelName(): string {
     return DEFAULT_MODEL
   }

packages/kilo-vscode/src/services/autocomplete/__tests__/AutocompleteModel.spec.ts

@@ -75,13 +75,6 @@ describe("AutocompleteModel", () => {
     })
   })
 
-  describe("supportsFim", () => {
-    it("always returns true", () => {
-      const model = new AutocompleteModel()
-      expect(model.supportsFim()).toBe(true)
-    })
-  })
-
   describe("getModelName", () => {
     it("returns the default model", () => {
       const model = new AutocompleteModel()
@@ -165,13 +158,4 @@ describe("AutocompleteModel", () => {
       )
     })
   })
-
-  describe("generateResponse", () => {
-    it("throws because FIM is the primary strategy", async () => {
-      const model = new AutocompleteModel()
-      await expect(model.generateResponse("system", "user", vi.fn())).rejects.toThrow(
-        "Chat-based completions are not supported via CLI backend",
-      )
-    })
-  })
 })

packages/kilo-vscode/src/services/autocomplete/classic-auto-complete/__tests__/AutocompleteInlineCompletionProvider.test.ts

@@ -851,7 +851,6 @@ describe("AutocompleteInlineCompletionProvider", () => {
       }),
       getModelName: vi.fn().mockReturnValue("test-model"),
       getProviderDisplayName: vi.fn().mockReturnValue("test-provider"),
-      supportsFim: vi.fn().mockReturnValue(true),
       hasValidCredentials: vi.fn().mockReturnValue(true), // Default to true for tests
     } as unknown as AutocompleteModel
     mockCostTrackingCallback = vi.fn() as CostTrackingCallback