test(connector-id): migrate fixtures and assertions to pi-mono ids

Replaces legacy adapter ids (codex_local / claude_local / gemini_local /
claude_compatible) across ~42 test files with pi-mono equivalents:
- codex_local → openai-codex
- claude_local → claude-pro
- gemini_local → gemini-cli
- claude_compatible → anthropic

Hand-rewrites where fixture semantics changed:
- connector-matching: TASK_TYPE_PREFERENCE ordering rebased (claude-pro
  first for code/content/management; google first for research)
- connector-model-policy: getModels(provider) from pi-ai instead of the
  deleted adapter.module.models array
- session-management + edge-case-session: adapter exemption removed;
  rotation now uniform under pi-mono
- ipc-core-admin-handlers: CSV compat model replaced with pi-ai anthropic
- connectors.spec: lookup by openai-codex id (catalog now starts with claude-pro)
- onboarding-model-pool: listProviderModels(provider) helper; label text
  changed to "ChatGPT Plus/Pro (Codex)"
- database integration: added vault insecure-fallback env + electron mock
- command-paths: uses new @main/connector-helpers normalizeCommandValue
- pi-mono-migrations: added `as string` casts for 3 negative-existence
  assertions verifying legacy ids were removed

Drops:
- pnpm-workspace.yaml packages: [] (was "packages/*"; directory empty)
- obsolete CONNECTOR_STRENGTHS import; .module.models probing

Verification:
- pnpm typecheck → 0 errors
- pnpm lint → 0 errors / 0 warnings
- pnpm build:verify → 0 errors / 0 warnings
- pnpm test:unit → 508/508 pass
- pnpm test:integration → 96/96 pass
- All 6 adapter hard gates pass (NOT FOUND / 0 matches / empty)
- pi-ai / pi-agent-core / pi-coding-agent all resolved @0.67.6

src/main/model-resolver.ts

@@ -0,0 +1,91 @@
+/**
+ * Resolves `Model<Api>` and the dynamic stream headers needed for providers
+ * that pi-ai's catalog does not ship out of the box.
+ *
+ * Today this only covers the two Fireworks connectors (`fireworks` and
+ * `fireworks-firepass`), which both speak the Anthropic-compatible endpoint
+ * at `https://api.fireworks.ai/inference` and require Bearer auth. For every
+ * other provider we delegate to pi-ai's `getModel` / `getModels`.
+ */
+
+import { getModel, getModels, type Api, type Model } from "@mariozechner/pi-ai";
+
+const FIREWORKS_BASE_URL = "https://api.fireworks.ai/inference";
+const FIREWORKS_DEFAULT_CONTEXT = 128_000;
+const FIREWORKS_DEFAULT_MAX_TOKENS = 8192;
+
+export const FIREWORKS_FIREPASS_MODEL_ID = "accounts/fireworks/routers/kimi-k2p5-turbo";
+export const FIREWORKS_FIREPASS_MODEL_NAME = "Kimi K2.5 Turbo (Fire Pass)";
+
+const FIREWORKS_CURATED_MODELS: Array<{ id: string; name: string; contextWindow?: number }> = [
+  { id: "accounts/fireworks/models/kimi-k2-instruct", name: "Kimi K2 Instruct", contextWindow: 131_072 },
+  { id: "accounts/fireworks/models/qwen3-coder-480b-a35b-instruct", name: "Qwen3 Coder 480B A35B Instruct", contextWindow: 262_144 },
+  { id: "accounts/fireworks/models/deepseek-v3p1", name: "DeepSeek V3.1", contextWindow: 163_840 },
+  { id: "accounts/fireworks/models/glm-4p5", name: "GLM 4.5", contextWindow: 131_072 },
+  { id: "accounts/fireworks/models/llama-v3p3-70b-instruct", name: "Llama 3.3 70B Instruct", contextWindow: 131_072 },
+];
+
+function buildFireworksModel(modelId: string, name?: string, contextWindow?: number): Model<"anthropic-messages"> {
+  return {
+    id: modelId,
+    name: name ?? modelId,
+    api: "anthropic-messages",
+    provider: "fireworks",
+    baseUrl: FIREWORKS_BASE_URL,
+    reasoning: false,
+    input: ["text"],
+    cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+    contextWindow: contextWindow ?? FIREWORKS_DEFAULT_CONTEXT,
+    maxTokens: FIREWORKS_DEFAULT_MAX_TOKENS,
+  };
+}
+
+function isFireworksProvider(provider: string): boolean {
+  return provider === "fireworks" || provider === "fireworks-firepass";
+}
+
+export function resolveModel(provider: string, modelId: string): Model<Api> | null {
+  if (provider === "fireworks-firepass") {
+    return buildFireworksModel(FIREWORKS_FIREPASS_MODEL_ID, FIREWORKS_FIREPASS_MODEL_NAME, 262_144);
+  }
+  if (provider === "fireworks") {
+    const curated = FIREWORKS_CURATED_MODELS.find((entry) => entry.id === modelId);
+    return buildFireworksModel(modelId, curated?.name, curated?.contextWindow);
+  }
+  try {
+    return getModel(provider as never, modelId as never) ?? null;
+  } catch {
+    return null;
+  }
+}
+
+export function resolveModels(provider: string): Array<{ id: string; name: string }> {
+  if (provider === "fireworks-firepass") {
+    return [{ id: FIREWORKS_FIREPASS_MODEL_ID, name: FIREWORKS_FIREPASS_MODEL_NAME }];
+  }
+  if (provider === "fireworks") {
+    return FIREWORKS_CURATED_MODELS.map(({ id, name }) => ({ id, name }));
+  }
+  try {
+    const models = getModels(provider as never) ?? [];
+    return models.map((model: { id: string; name?: string }) => ({ id: model.id, name: model.name ?? model.id }));
+  } catch {
+    return [];
+  }
+}
+
+/**
+ * Extra HTTP headers that must be merged into the streamSimple call for the
+ * given provider. Returns undefined when no extra headers are required.
+ *
+ * Fireworks' Anthropic-compat endpoint authenticates via `Authorization: Bearer`,
+ * not `x-api-key`. pi-ai's anthropic provider only sends `x-api-key`, so we
+ * inject the bearer header at the call site.
+ */
+export function resolveExtraStreamHeaders(provider: string, apiKey: string): Record<string, string> | undefined {
+  if (!apiKey) return undefined;
+  if (isFireworksProvider(provider)) {
+    return { Authorization: `Bearer ${apiKey}` };
+  }
+  return undefined;
+}