Skip to content
This repository was archived by the owner on Oct 16, 2026. It is now read-only.
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions apps/cli/tsup.config.ts
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@ export default defineConfig({
'@github/copilot-sdk',
'@openai/codex-sdk',
'@anthropic-ai/claude-agent-sdk',
'ai',
'@ai-sdk/openai',
],
// Provide a real require() for bundled CJS modules (e.g. debug) that need Node.js builtins like tty
banner: {
Expand Down
40 changes: 40 additions & 0 deletions apps/web/src/content/docs/docs/targets/coding-agents.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -227,6 +227,46 @@ targets:
grader_target: azure-base
```

## AI SDK Agent

`provider: ai-sdk-agent` is experimental. It uses Vercel AI SDK model and
tool-calling primitives with AgentV-owned coding tools and a bounded tool loop.
It is not a one-shot LLM target.

```yaml
targets:
- name: ai-sdk-agent
provider: ai-sdk-agent
base_url: ${{ AGENTV_OPENAI_BASE_URL }}
api_key: ${{ AGENTV_OPENAI_API_KEY }}
model: ${{ AGENTV_CODEX_MODEL }}
max_steps: 20
tools: read,bash,edit,write
grader_target: azure-base
```

| Field | Required | Description |
|-------|----------|-------------|
| `base_url` | No | OpenAI-compatible base URL. Defaults to `https://api.openai.com/v1`. |
| `api_key` | Yes | OpenAI-compatible API key. Prefer `${{ ENV_VAR }}` references. |
| `model` | Yes | Chat-completions model id. |
| `temperature` | No | Model temperature. |
| `max_steps` | No | Agent loop step limit. Defaults to `20`. |
| `tools` | No | Comma-separated allowlist from `read,bash,edit,write`. Defaults to all four. |
| `system_prompt` | No | Extra system instructions appended to AgentV's coding-agent prompt. |
| `cwd` | No | Working directory override. |
| `timeout_seconds` | No | Per-case model-call timeout. |
| `grader_target` | Yes | LLM target for evaluation. |

The first slice includes minimal repo-owned tools: `read` and `write` operate on
UTF-8 files relative to the eval workspace, `edit` performs exact string
replacement, and `bash` runs shell commands in the workspace with timeout and
bounded stdout/stderr capture.

AgentV loads `ai` and `@ai-sdk/openai` only when this provider is invoked. If
they are not installed locally, AgentV uses a managed dependency directory under
`~/.agentv/deps/ai-sdk-agent`.

## VS Code

```yaml
Expand Down
1 change: 1 addition & 0 deletions apps/web/src/content/docs/docs/targets/configuration.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,7 @@ already-exported secrets into `.env`.
| `gemini` | LLM | Google Gemini |
| `claude` | Agent | Claude Agent SDK |
| `codex` | Agent | Codex CLI |
| `ai-sdk-agent` | Agent | Experimental Vercel AI SDK coding-agent harness |
| `pi-coding-agent` | Agent | Pi Coding Agent |
| `vscode` | Agent | VS Code with Copilot |
| `vscode-insiders` | Agent | VS Code Insiders |
Expand Down
Loading
Loading