uncaged/nerve
Archived
10
0
Fork 0
Code Issues 5 Pull Requests Actions Packages Projects Releases Wiki Activity

feat(workflow-utils): add createLlmAdapter AgentFn factory #278

Merged
xingyue merged 2 commits from refactor/277-llm-adapter-four-tuple into main 2026-04-30 12:51:30 +00:00
Conversation 0 Commits 2 Files Changed 4
+79 -19
xiaoju commented 2026-04-30 12:42:04 +00:00
Owner
Copy Link
Copy Source

What

Add createLlmAdapter(provider): AgentFn — a single-turn LLM chat adapter that conforms to the AgentFn interface, enabling pure-LLM roles to use the standard createRole(adapter, prompt, schema, extract) four-tuple pattern.

Why

questioner and answerer roles in knowledge-extraction were hand-assembled instead of using createRole. This adapter fills the gap — roles that don't need an agent (cursor/hermes) can use a simple LLM adapter instead.

Changes

  • create-llm-adapter.ts — new file: createLlmAdapter(provider) returns AgentFn using chatCompletionText
  • index.ts — export createLlmAdapter
  • create-llm-adapter.test.ts — unit tests

Workspace-side changes (questioner/answerer refactor to use createRole) committed separately to nerve-workspace.

Ref

Fixes #277

## What Add `createLlmAdapter(provider): AgentFn` — a single-turn LLM chat adapter that conforms to the `AgentFn` interface, enabling pure-LLM roles to use the standard `createRole(adapter, prompt, schema, extract)` four-tuple pattern. ## Why `questioner` and `answerer` roles in knowledge-extraction were hand-assembled instead of using `createRole`. This adapter fills the gap — roles that don't need an agent (cursor/hermes) can use a simple LLM adapter instead. ## Changes - **`create-llm-adapter.ts`** — new file: `createLlmAdapter(provider)` returns `AgentFn` using `chatCompletionText` - **`index.ts`** — export `createLlmAdapter` - **`create-llm-adapter.test.ts`** — unit tests Workspace-side changes (questioner/answerer refactor to use `createRole`) committed separately to `nerve-workspace`. ## Ref Fixes #277
xiaoju added 1 commit 2026-04-30 12:42:05 +00:00
feat(workflow-utils): add createLlmAdapter AgentFn factory fbe1cc8eba 
Single-turn chat via chatCompletionText: system from createRole prompt, user from ctx.start.content.

Fixes #277

Made-with: Cursor
xiaoju added 1 commit 2026-04-30 12:47:14 +00:00
chore(workflow-utils): remove deprecated role factory exports 005739f6bc 
Remove createCursorRole, createHermesRole, createLlmRole, createReActRole
from public API — all superseded by createRole(adapter, prompt, schema, extract).
Source files retained as internal implementation.

Also remove unused type exports: CliPromptFn, CursorRole*, HermesRole*,
LlmPromptFn, LlmRole*, ReActRole*, ReActTool. Keep LlmMessage and
MetaExtractConfig (used internally).

Refs #277
xingyue merged commit d2bb0275dc into main 2026-04-30 12:51:30 +00:00
This repo is archived. You cannot comment on pull requests.
Reviewers
No Reviewers
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
aobing jiashuang (Jia Shuang) jiayi (Jiayi) luming scottwei tuanzi xiaoju xiaomo xiaonuo xingyue
No Assignees
1 Participants
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: uncaged/nerve#278
Delete Branch "refactor/277-llm-adapter-four-tuple"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?