Skip to content

Releases: open-runtime/opencode

v1.1.50-open-runtime-fork.3

07 Feb 00:50

Choose a tag to compare

fix(plugin): remove unnecessary default auth plugins from BUILTIN

These third-party auth plugins (opencode-anthropic-auth, gitlab-auth) are
unnecessary for our embedded runtime use case where auth is handled by the
host Dart process. They cause intermittent ERR_MODULE_NOT_FOUND failures
when Bun's compiled binary can't resolve npm packages installed at runtime
from /$bunfs/, and add ~1.5s of startup latency per execution.

Co-authored-by: Cursor <cursoragent@cursor.com>

v1.1.50-open-runtime-fork.2

06 Feb 20:27

Choose a tag to compare

fix(plugin): use neutral agent for RuntimeAPI to prevent prompt conta…

v1.1.50-open-runtime-fork.1

06 Feb 19:44

Choose a tag to compare

fix(plugin): fix RuntimeAPI getProviders, getModels, and stream error…

v1.1.48-open-runtime-fork.6

04 Feb 05:19

Choose a tag to compare

fix: extract base semver for plugin version matching

Handles custom fork version strings like "1.1.48-open-runtime-fork.5"
by extracting the base semver "1.1.48" for @opencode-ai/plugin installation.

This allows the fork to use the official @opencode-ai/plugin package
from npm without requiring a separately published fork plugin.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

v1.1.48-open-runtime-fork.5

04 Feb 02:50

Choose a tag to compare

fix(plugin): Add complete model structure to RuntimeAPI fallback

The fallback model was missing required fields like `capabilities`,
`cost`, `limit`, etc. that LLM.stream() expects. This caused errors like
"undefined is not an object (evaluating 'input.model.capabilities.temperature')".

Now creates a complete model object with sensible defaults for all
required fields, enabling RuntimeAPI to work when Provider.getModel()
fails (e.g., for providers not in models.dev like OpenRouter).

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

v1.1.48-open-runtime-fork.4

04 Feb 02:42

Choose a tag to compare

fix(plugin): Include api.npm in RuntimeAPI fallback model

When Provider.getModel() fails, the RuntimeAPI creates a fallback model
object. This fallback was missing the api.npm field required by getSDK()
to load the correct provider SDK.

Added a mapping of known provider IDs to their npm packages, with a
fallback to @ai-sdk/openai-compatible for unknown providers.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

v1.1.48-open-runtime-fork.3

04 Feb 02:23

Choose a tag to compare

feat: expose RuntimeAPI to plugins for LLM completion calls

Adds a new RuntimeAPI that plugins can use to make LLM completion calls
without disrupting the active session. The API provides:
- stream(): Stream LLM completions with deltas
- getProviders(): List available providers
- getModels(): List available models for a provider

This allows the Dart runtime plugin to make LLM calls through the
official plugin API rather than relying on internal module imports.

v1.1.48-open-runtime-fork.2

04 Feb 02:04

Choose a tag to compare

Rename storage directory from 'opencode' to 'execution'

v1.1.48-open-runtime-fork.1

04 Feb 01:54

Choose a tag to compare

Fix archive path: ../../ not ../../../ from bin directory