feat(ollama): return real token counts from embedding endpoint

[pyramation]

Summary

Ollama's /api/embed endpoint returns prompt_eval_count (input token count) but agentic-kit was discarding it. This PR:

1. Switches from deprecated /api/embeddings to /api/embed — the newer endpoint supports batch input and returns prompt_eval_count
2. generateEmbedding() now returns EmbeddingResult instead of number[]:

   interface EmbeddingResult {
     embedding: number[];
     promptTokens: number;
   }

3. Adds OllamaAdapter.embed() convenience method that delegates to the client

BREAKING CHANGE: generateEmbedding() return type changed from Promise<number[]> to Promise<EmbeddingResult>. Callers need to use result.embedding instead of the raw array.

Before:

const embedding = await client.generateEmbedding('hello');
// embedding is number[] — no token info

After:

const result = await client.generateEmbedding('hello');
// result.embedding is number[], result.promptTokens is the real count

This unblocks constructive's metering layer from using real token counts for embeddings (currently using ~4 chars/token placeholder).

Review & Testing Checklist for Human

• Run live tests against a real Ollama instance with nomic-embed-text installed: OLLAMA_LIVE_MODEL=qwen3.5:4b pnpm --filter @agentic-kit/ollama test:live:extended — verify promptTokens > 0 in the new embedding tests
• Verify the /api/embed endpoint works with your Ollama version (requires Ollama ≥ 0.4.0 for /api/embed; older versions only have /api/embeddings)
• Check downstream consumers that call generateEmbedding() — they need to update from result (array) to result.embedding (array) and can now access result.promptTokens

Notes

• The pre-existing agentic-kit package test failure (reasoning field mismatch with ProviderAdapter type) is unrelated — same failure on main
• The old /api/embeddings endpoint (singular) returned { embedding: number[] }. The new /api/embed returns { embeddings: number[][], prompt_eval_count: number } — we take embeddings[0] for single-text input

Link to Devin session: https://app.devin.ai/sessions/2b5a29d83d3f478e8d3d972653b4879c
Requested by: @pyramation

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring