Files
PaperClipAI/packages/adapters/opencode-local/src/index.ts
T
Devin FoleyandPaperclip cdf04a33fa feat(adapters): refresh current coding models and reasoning controls (#13829)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - Its adapters supply model catalogs and reasoning controls to agent
setup.
> - Several provider releases are missing from the fallback catalogs.
> - Some newer models also have effort levels that the UI does not
offer.
> - Operators need the exact supported IDs and controls when discovery
is unavailable.
> - This pull request updates the existing adapters from current
provider documentation.
> - Operators can select current coding models without entering custom
IDs.

## Linked Issues or Issue Description

**What existing behavior does this improve?**

Model selection and reasoning controls across the existing coding-agent
adapters.

**Subsystem affected**

Claude, Codex, Grok, Gemini, Cursor, Kimi, and OpenCode adapters; model
discovery tests; agent creation and editing.

**Current behavior**

The catalogs omit Opus 5.5, GPT-6 Sol/Luna, Grok 4.7/4.6/4.5, current
Gemini Flash models, and several Cursor/Kimi choices. Bedrock has
obsolete IDs. The UI omits supported effort levels and saves Grok effort
under a key the runtime does not read.

**Proposed behavior**

Offer verified current model IDs and model-specific efforts. Remove
retired Gemini 2.0 choices. Keep configured defaults and saved model
IDs. Keep runtime discovery for account-specific choices.

**Reason and benefit**

Catch up with provider releases through September 22, 2026. Correct the
picker and runtime controls together.

**Breaking changes**

No database or API change. Gemini 2.0 options leave the picker after
their June 1 shutdown. Existing saved IDs remain unchanged. Corrected
Bedrock catalog IDs do not rewrite saved configuration.

**Additional context**

Supersedes the separate GPT-6 Sol PR #13830. Fable 5.1 was already
merged in #12730, and GPT-6 Astra in #12851. The Grok 4.6/4.5 proposal
#11324 was closed and parked by its author. This change retains the
default-sentinel fix from #12062. Related discovery proposals #13127 and
#13565 do not supply these catalog and effort updates. Searches found no
open PR for the additional model IDs.

See [the dated
audit](https://github.com/paperclipai/paperclip/blob/feat/claude-opus-5-5/doc/adapter-model-audit-2026-09-22.md)
for exact scope, primary sources, runtime observations, and
account-specific limits. This updates existing adapters and does not
duplicate planned core work.

## What Changed

- Add Opus 5.5 for direct Claude and Bedrock, with a Claude Code 2.1.280
gate. Correct and extend Bedrock model IDs.
- Add GPT-6 Sol/Luna and Fast mode. Offer Ultra for Astra/Sol and
GPT-5.6 Sol/Terra, and Max for both Luna generations.
- Add Grok 4.7/4.6/4.5, expose supported Extra High effort, and save
Grok edits under `reasoningEffort`.
- Add Gemini Flash 3.8/3.7/3.6/3.5, Flash Lite 3.5/3.1, and 3 Flash
Preview. Remove retired 2.0 choices.
- Add the current documented Cursor fallback models, including Fable
5.1, Composer 2.5, and Muse Spark 1.3.
- Refresh OpenCode fallback IDs used in remote environments from its
installed provider registry.
- Add Kimi K3 256K. Update the existing coding alias to K2.8 Preview and
enable its CLI effort settings.
- Use model-specific Claude/Grok efforts in creation and editing. Clear
unsupported effort when switching models.
- Add catalog, CLI/ACP forwarding, compatibility, and UI persistence
coverage. Record the audit and sources.

## Verification

- Latest head `6e63c9ef53b54ba869cd4fb431a8570bebe289f4`: 53 CI checks
passed, 2 skipped. This includes full workspace typecheck, build, and
all test shards. Greptile is 5/5 with zero unresolved threads. GitHub
reports no merge conflicts.
- 340 focused tests passed across adapter metadata, CLI/ACP arguments,
Claude version checks, Kimi effort, Grok execution, server model
discovery, and UI effort selection/persistence.
- `pnpm --filter @paperclipai/adapter-claude-local --filter
@paperclipai/adapter-codex-local --filter
@paperclipai/adapter-grok-local --filter
@paperclipai/adapter-gemini-local --filter
@paperclipai/adapter-kimi-local --filter
@paperclipai/adapter-cursor-local --filter
@paperclipai/adapter-opencode-local typecheck` — passed. The same
filters with `build` passed.
- `pnpm check:token-gates` and `git diff --check` — passed.
- Full workspace and UI typechecks were attempted locally. They stop on
existing missing `three` dependencies in `packages/shared/src/cliplab`.
- Full `pnpm test:run` and `pnpm build` were not run locally. Worktree
creation exhausted disk space, so a clean dependency install is not
feasible on this host. Focused checks reuse existing dependencies. CI
supplies full workspace verification.
- No provider inference was run. Account-specific runtime model lists
were inspected where available.
- Manual check: select the new models in agent setup and editing.
Confirm Luna has Max but no Ultra, Grok 4.7 has Extra High, and Fable
5.1 has Extra High/Max. Save Grok effort and confirm
`adapterConfig.reasoningEffort` contains the selection.

## Risks

- Catalog presence does not grant account access. Older CLIs and
restricted accounts can reject a model. Opus 5.5 has an explicit upgrade
check.
- Higher effort can increase cost and latency. Existing agent defaults
are unchanged.
- Cursor fallback IDs come from public model documentation; the local
account exposed no live catalog. Runtime discovery still adds
account-specific variants.
- Kimi effort remains supported only on its explicit CLI engine. This
does not add effort support to its default ACP engine.
- Saved obsolete Bedrock or retired Gemini IDs are not migrated
automatically.

## Model Used

OpenAI GPT-6 through Codex, with reasoning, tool use, and code
execution. The exact deployment ID and context window are not exposed to
this session.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-09-22 16:56:00 -07:00

118 lines
6.4 KiB
TypeScript

export const type = "opencode_local";
export const label = "OpenCode";
// Use OpenCode's official installer instead of `npm install -g opencode-ai`.
// The npm package reifies four large Linux x64 prebuilt-binary subpackages
// (linux-x64, linux-x64-musl, linux-x64-baseline, linux-x64-baseline-musl) in
// parallel even though only one matches the sandbox; on bandwidth-constrained
// sandboxes (e.g. Cloudflare) that exceeded the 240s install budget. The
// official installer fetches a single arch-specific binary into
// `$HOME/.opencode/bin` and tries to add it to PATH via `~/.bashrc`. That
// rc-file path is only sourced by interactive/login shells, so non-login
// `sh -c` probe invocations (used by the runtime PATH check) cannot find the
// binary. We fix that by symlinking the installed binary into a directory on
// the non-login `sh -c` PATH: prefer `/usr/local/bin` (universally on the
// default PATH on Linux distros) when root or passwordless sudo is available,
// otherwise fall back to `$HOME/.local/bin` (which is on the default PATH on
// the exe.dev sandbox image and most modern home-managed Linux images).
//
// Security tradeoff: this is `curl | bash` without a SHA-256 verification of
// the install script. We accept this because:
// 1. The install runs inside an isolated, ephemeral sandbox — blast radius
// is bounded to that sandbox's secrets and disk.
// 2. The prior `npm install -g opencode-ai` is also unverified code
// execution from a third-party registry; this is not strictly worse.
// 3. OpenCode does not publish per-release SHA-256 checksums in a stable
// location, and pinning a version + hash here would require manual
// version bumps on every OpenCode release.
// The `set -e` (implied by Bash's default with `-fsSL` upstream of a piped
// shell) and `curl -fsSL` give us fail-fast behavior on HTTP errors. If
// OpenCode starts publishing a stable checksum/signature, switch to fetching
// a versioned tarball + verifying the digest before exec.
export const SANDBOX_INSTALL_COMMAND =
'curl -fsSL https://opencode.ai/install | bash && ' +
'if [ -x "$HOME/.opencode/bin/opencode" ]; then ' +
'if [ "$(id -u)" -eq 0 ]; then ' +
'ln -sf "$HOME/.opencode/bin/opencode" /usr/local/bin/opencode; ' +
'elif command -v sudo >/dev/null 2>&1 && sudo -n true >/dev/null 2>&1; then ' +
'sudo ln -sf "$HOME/.opencode/bin/opencode" /usr/local/bin/opencode; ' +
'else ' +
'mkdir -p "$HOME/.local/bin" && ' +
'ln -sf "$HOME/.opencode/bin/opencode" "$HOME/.local/bin/opencode"; ' +
'fi; ' +
'fi';
export const DEFAULT_OPENCODE_LOCAL_MODEL = "openai/gpt-5.2-codex";
export function isValidOpenCodeModelId(value: unknown): value is string {
if (typeof value !== "string") return false;
const trimmed = value.trim();
const slashIndex = trimmed.indexOf("/");
return Boolean(trimmed) && slashIndex > 0 && slashIndex !== trimmed.length - 1;
}
export const models: Array<{ id: string; label: string }> = [
{ id: DEFAULT_OPENCODE_LOCAL_MODEL, label: DEFAULT_OPENCODE_LOCAL_MODEL },
{ id: "openai/gpt-6-astra", label: "openai/gpt-6-astra" },
{ id: "openai/gpt-6-sol", label: "openai/gpt-6-sol" },
{ id: "openai/gpt-6-luna", label: "openai/gpt-6-luna" },
{ id: "openai/gpt-5.6-sol", label: "openai/gpt-5.6-sol" },
{ id: "openai/gpt-5.6-terra", label: "openai/gpt-5.6-terra" },
{ id: "openai/gpt-5.6-luna", label: "openai/gpt-5.6-luna" },
{ id: "anthropic/claude-opus-5-5", label: "anthropic/claude-opus-5-5" },
{ id: "anthropic/claude-opus-5", label: "anthropic/claude-opus-5" },
{ id: "anthropic/claude-fable-5-1", label: "anthropic/claude-fable-5-1" },
{ id: "anthropic/claude-sonnet-5", label: "anthropic/claude-sonnet-5" },
{ id: "google/gemini-3.8-flash", label: "google/gemini-3.8-flash" },
{ id: "xai/grok-4.7", label: "xai/grok-4.7" },
{ id: "openai/gpt-5.5", label: "openai/gpt-5.5" },
{ id: "openai/gpt-5.4", label: "openai/gpt-5.4" },
{ id: "openai/gpt-5.4-mini", label: "openai/gpt-5.4-mini" },
{ id: "openai/gpt-5.2", label: "openai/gpt-5.2" },
{ id: "openai/gpt-5.1-codex-max", label: "openai/gpt-5.1-codex-max" },
{ id: "openai/gpt-5.1-codex-mini", label: "openai/gpt-5.1-codex-mini" },
];
export const agentConfigurationDoc = `# opencode_local agent configuration
Adapter: opencode_local
Use when:
- You want Paperclip to run OpenCode locally as the agent runtime
- You want provider/model routing in OpenCode format (provider/model)
- You want OpenCode session resume across heartbeats via --session
Don't use when:
- You need webhook-style external invocation (use openclaw_gateway or http)
- You only need one-shot shell commands (use process)
- OpenCode CLI is not installed on the machine
Core fields:
- cwd (string, optional): default absolute working directory fallback for the agent process (created if missing when possible)
- instructionsFilePath (string, optional): absolute path to a markdown instructions file prepended to the run prompt
- model (string, required): OpenCode model id in provider/model format (for example anthropic/claude-sonnet-4-5)
- variant (string, optional): provider-specific reasoning/profile variant passed as --variant (for example minimal|low|medium|high|xhigh|max)
- dangerouslySkipPermissions (boolean, optional): inject a runtime OpenCode config with \`permission=allow\` for all tools and connections; defaults to true for unattended Paperclip runs
- promptTemplate (string, optional): run prompt template
- command (string, optional): defaults to "opencode"
- extraArgs (string[], optional): additional CLI args
- env (object, optional): KEY=VALUE environment variables
Operational fields:
- timeoutSec (number, optional): run timeout in seconds
- graceSec (number, optional): SIGTERM grace period in seconds
Notes:
- OpenCode supports multiple providers and models. Use \
\`opencode models\` to list available options in provider/model format.
- Paperclip requires an explicit \`model\` value for \`opencode_local\` agents.
- Runs are executed with: opencode run --format json ...
- Sessions are resumed with --session when stored session cwd matches current cwd.
- The adapter sets OPENCODE_DISABLE_PROJECT_CONFIG=true to prevent OpenCode from \
writing an opencode.json config file into the project working directory. Model \
selection is passed via the --model CLI flag instead.
- When \`dangerouslySkipPermissions\` is enabled, Paperclip injects a temporary \
runtime config with \`permission=allow\` so headless runs do \
not stall on approval prompts.
`;