feat: 8sync opencode connect gguf + 8sync gsd connect gguf

[8syncdev]

Summary

Two new commands to register a running llama-server as a local provider in GSD and OpenCode, with correct provider schemas.

8sync gsd connect gguf (alias for add gguf)

Probes llama-server on localhost, writes entry to ~/.gsd/agent/models.json.

8sync gsd connect gguf              # probe ports 8080/8081/8082/1234/11434
8sync gsd connect gguf --port 8081  # specific port
8sync gsd connect gguf --name my-id # custom provider id
8sync gsd connect gguf --dry-run    # preview only

Fix: Missing apiKey field caused GSD to reject the models.json entry with:

Provider gguf-local-model: "apiKey" is required when defining custom models.

Added apiKey: "local" (GSD schema requires the field; llama-server ignores it).

8sync opencode connect gguf

Probes llama-server, patches ~/.config/opencode/opencode.json.

8sync opencode connect gguf
8sync opencode connect gguf --port 8081
8sync opencode connect gguf --name my-local --dry-run

Fix: OpenCode schema requires:

• options.apiKey + options.baseURL (not top-level apiKey)
• models as a key-value object, not an array

Previous attempt caused:

Invalid input: expected record, received array provider.gguf-local-model.models
Unrecognized key: "apiKey" provider.gguf-local-model

Correct structure written:

"gguf-local-model": {
  "options": { "apiKey": "local", "baseURL": "http://localhost:8080/v1" },
  "models": {
    "model.gguf": { "name": "model (local GGUF)", "attachment": false, "reasoning": false }
  }
}

Schema verified against https://opencode.ai/config.json.

Full workflow

8sync gguf serve --engine-path "C:\...\run" --model-path "C:\...\model.gguf" --balance
8sync gsd connect gguf        # registers in models.json
8sync opencode connect gguf   # registers in opencode.json
# /model -> select gguf-local-model/model.gguf
8sync gguf stop               # when done

Commit

5dfea7e