`js_repl_tools_only` can be enabled to force direct model tool calls through `js_repl`:
```toml
[features]
js_repl = true
js_repl_tools_only = true
```
When enabled, direct model tool calls are restricted to `js_repl` and `js_repl_reset`; other tools remain available via `await codex.tool(...)` inside js_repl.
2.`js_repl_node_module_dirs` in config/profile (array of absolute paths)
3. Thread working directory (cwd, always included as the last fallback)
For `CODEX_JS_REPL_NODE_MODULE_DIRS` and `js_repl_node_module_dirs`, module resolution is attempted in the order provided with earlier entries taking precedence.
- If a cell throws, prior bindings remain available, lexical bindings whose initialization completed before the throw stay available in later calls, and hoisted `var` / `function` bindings persist only when execution clearly reached their declaration or a supported write site.
- Supported hoisted-`var` failed-cell cases are direct top-level identifier writes and updates before the declaration (for example `x = 1`, `x += 1`, `x++`, `x &&= 1`) and non-empty top-level `for...in` / `for...of` loops.
- Intentionally unsupported failed-cell cases include hoisted function reads before the declaration, aliasing or direct-IIFE-based inference, writes in nested blocks or other nested statement structure, nested writes inside already-instrumented assignment RHS expressions, destructuring-assignment recovery for hoisted `var`, partial `var` destructuring recovery, pre-declaration `undefined` reads, and empty top-level `for...in` / `for...of` loop vars.
- Imported local files must be ESM `.js` / `.mjs` files and run in the same REPL VM context as the calling cell.
- Static imports inside imported local files may only target other local `.js` / `.mjs` files via relative paths, absolute paths, or `file://` URLs. Bare package and builtin imports from local files must stay dynamic via `await import(...)`.
-`import.meta.resolve()` returns importable strings such as `file://...`, bare package names, and `node:fs`; the returned value can be passed back to `await import(...)`.
- Local file modules reload between execs, so a later `await import("./file.js")` picks up edits and fixed failures. Top-level bindings you already created still persist until `js_repl_reset`.
-`codex.tool(name, args?)`: executes a normal Codex tool call from inside `js_repl` (including shell tools like `shell` / `shell_command` when available).
-`codex.tool(...)` and `codex.emitImage(...)` keep stable helper identities across cells. Saved references and persisted objects can reuse them in later cells, but async callbacks that fire after a cell finishes still fail because no exec is active.
- Each `codex.tool(...)` call emits a bounded summary at `info` level from the `codex_core::tools::js_repl` logger. At `trace` level, the same path also logs the exact raw response object or error string seen by JavaScript.
-`codex.emitImage(...)` accepts a data URL, a single `input_image` item, an object like `{ bytes, mimeType }`, or a raw tool response object that contains exactly one image and no text. Call it multiple times if you want to emit multiple images.
- Request full-resolution image processing with `detail: "original"` only when the `view_image` tool schema includes a `detail` argument. The same availability applies to `codex.emitImage(...)`: if `view_image.detail` is present, you may also pass `detail: "original"` there. Use this when high-fidelity image perception or precise localization is needed, especially for CUA agents.
- Example of sharing an in-memory Playwright screenshot: `await codex.emitImage({ bytes: await page.screenshot({ type: "jpeg", quality: 85 }), mimeType: "image/jpeg", detail: "original" })`.
- Example of sharing a local image tool result: `await codex.emitImage(codex.tool("view_image", { path: "/absolute/path", detail: "original" }))`.
- When encoding an image to send with `codex.emitImage(...)` or `view_image`, prefer JPEG at about 85 quality when lossy compression is acceptable; use PNG when transparency or lossless detail matters. Smaller uploads are faster and less likely to hit size limits.
