HarvmasterandClaude Opus 5.5
fa8f775779
- models/: GGUFs (Git LFS), manifest.json, model card, Apache-2.0 LICENSE + NOTICE - extensions/vscode/: the Local Autocomplete extension (v0.5.0, publisher harvmaster) - daemon/: acd completion daemon (bundled into the extension; also runs standalone) - shared/: prompt layout + symbol extraction shared with the training pipeline - scripts/: make-model-release.mjs, publish-extension-release.mjs (Gitea releases) - README with quick start, settings, architecture, build and release instructions; MIT license for code; CHANGELOG The default model manifest URL moves to models/manifest.json. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
78 lines
3.1 KiB
TypeScript
78 lines
3.1 KiB
TypeScript
import fs from 'node:fs'
|
|
import { Tokenizer } from '@huggingface/tokenizers'
|
|
import { chunkedPrefixTokens, fimPrompt, headTokens, repoHeader, wantsSingleLine } from '../../shared/prompt.ts'
|
|
import type { ContextMode, DaemonConfig, ModelProfile } from './config.ts'
|
|
|
|
export interface FimRequest {
|
|
requestId?: string
|
|
/** Workspace folder name; Stage C was trained with the repo directory name here. */
|
|
repo: string
|
|
/** Path relative to the workspace folder, '/'-separated. */
|
|
path: string
|
|
/** Absolute workspace folder on the daemon's machine; enables symbols mode and the this.-repair. */
|
|
workspaceRoot?: string
|
|
languageId?: string
|
|
prefix: string
|
|
suffix: string
|
|
/** Absolute line number of prefix's first line, if the client sent a truncated document. */
|
|
prefixLineOffset?: number
|
|
mode?: ContextMode
|
|
/** Override automatic single-line detection. */
|
|
singleLine?: boolean
|
|
maxLines?: number
|
|
}
|
|
|
|
export interface BuiltPrompt {
|
|
ids: number[]
|
|
mode: ContextMode
|
|
singleLine: boolean
|
|
prefixStartLine: number
|
|
tokenizeMs: number
|
|
}
|
|
|
|
export const STOP_STRINGS = ['<|endoftext|>', '<|fim_prefix|>', '<|fim_suffix|>', '<|fim_middle|>', '<|fim_pad|>', '<|file_sep|>', '<|repo_name|>', '<|im_start|>', '<|im_end|>']
|
|
|
|
export class PromptBuilder {
|
|
private encodeFn: (s: string) => number[]
|
|
private fim: { prefix: number; suffix: number; middle: number }
|
|
private header: { repoName: number; fileSep: number }
|
|
private config: DaemonConfig
|
|
|
|
constructor(config: DaemonConfig) {
|
|
this.config = config
|
|
const tokenizer = new Tokenizer(
|
|
JSON.parse(fs.readFileSync(config.tokenizer.json, 'utf8')),
|
|
JSON.parse(fs.readFileSync(config.tokenizer.config, 'utf8')),
|
|
)
|
|
this.encodeFn = s => tokenizer.encode(s, { add_special_tokens: false }).ids
|
|
const id = (t: string) => {
|
|
const v = tokenizer.token_to_id(t)
|
|
if (v === undefined) throw new Error(`tokenizer is missing ${t}`)
|
|
return v
|
|
}
|
|
this.fim = { prefix: id('<|fim_prefix|>'), suffix: id('<|fim_suffix|>'), middle: id('<|fim_middle|>') }
|
|
this.header = { repoName: id('<|repo_name|>'), fileSep: id('<|file_sep|>') }
|
|
}
|
|
|
|
encode(s: string) { return this.encodeFn(s) }
|
|
|
|
/**
|
|
* Builds the prompt in the model's trained layout. `symbols` context is passed
|
|
* in by the caller (empty until the symbol provider exists, which degrades to `simple`).
|
|
*/
|
|
build(req: FimRequest, profile: ModelProfile, mode: ContextMode, symbols?: string): BuiltPrompt {
|
|
const t0 = performance.now()
|
|
const p = this.config.prompt
|
|
const { tokens: prefix, startLine } = chunkedPrefixTokens(req.prefix, p.prefixTokens, this.encodeFn, p.prefixChunkLines, req.prefixLineOffset ?? 0)
|
|
const suffix = headTokens(req.suffix, p.suffixTokens, this.encodeFn)
|
|
const header = mode === 'raw' ? [] : repoHeader(req.repo, req.path, mode === 'symbols' ? symbols : undefined, this.encodeFn, this.header)
|
|
return {
|
|
ids: [...header, ...fimPrompt(profile.order, prefix, suffix, this.fim)],
|
|
mode,
|
|
singleLine: req.singleLine ?? wantsSingleLine(req.suffix),
|
|
prefixStartLine: startLine,
|
|
tokenizeMs: performance.now() - t0,
|
|
}
|
|
}
|
|
}
|