pi-knowledge 0.6.0 → 0.8.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +16 -0
- package/README.md +3 -1
- package/dist/src/model-worker.js +16 -0
- package/dist/src/search/reranker-api.js +6 -6
- package/dist/src/search/reranker-config.d.ts +4 -1
- package/dist/src/search/reranker-config.js +11 -5
- package/dist/src/search/reranker-logits.d.ts +20 -0
- package/dist/src/search/reranker-logits.js +16 -0
- package/docs/configuration.md +7 -4
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,21 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
|
|
4
|
+
## [0.8.0] - 2026-08-04
|
|
5
|
+
|
|
6
|
+
### Added
|
|
7
|
+
- Added Hugging Face Text Embeddings Inference reranker API format support through `PI_KNOWLEDGE_RERANKER_API_FORMAT=tei`.
|
|
8
|
+
|
|
9
|
+
|
|
10
|
+
## [0.7.0] - 2026-07-28
|
|
11
|
+
|
|
12
|
+
### Added
|
|
13
|
+
- Added `PI_KNOWLEDGE_RERANKER_RAW_LOGITS` for trusted single-logit local/Hugging Face cross-encoder rerankers whose sigmoid scores flatten near 1.0.
|
|
14
|
+
|
|
15
|
+
### Fixed
|
|
16
|
+
- Preserved the default reranker config shape when raw logits are unset.
|
|
17
|
+
- Rejected multi-logit and non-finite raw reranker outputs instead of guessing a relevance score.
|
|
18
|
+
|
|
3
19
|
## [0.6.0] - 2026-07-25
|
|
4
20
|
|
|
5
21
|
### Added
|
package/README.md
CHANGED
|
@@ -179,11 +179,13 @@ Full configuration details are in [docs/configuration.md](docs/configuration.md)
|
|
|
179
179
|
| Storage path | `PI_KNOWLEDGE_DIR`, `OMP_KNOWLEDGE_DIR`, `PI_CODING_AGENT_DIR`, `OMP_CODING_AGENT_DIR`, `OMP_PROFILE` |
|
|
180
180
|
| Model worker and cache | `PI_KNOWLEDGE_MODEL_CACHE_DIR`, `PI_KNOWLEDGE_NODE_PATH` |
|
|
181
181
|
| Embedding provider | `PI_KNOWLEDGE_EMBEDDING`, `OPENAI_API_KEY`, `PI_KNOWLEDGE_EMBEDDING_BASE_URL`, `OPENAI_BASE_URL`, `PI_KNOWLEDGE_EMBEDDING_MAX_CHARS`, `PI_KNOWLEDGE_EMBEDDING_API_FALLBACK` |
|
|
182
|
-
| Reranker provider | `PI_KNOWLEDGE_RERANKER`, `PI_KNOWLEDGE_RERANKER_REVISION`, `PI_KNOWLEDGE_RERANKER_DTYPE`, `PI_KNOWLEDGE_RERANKER_REMOTE_HOST`, `PI_KNOWLEDGE_RERANKER_REMOTE_PATH_TEMPLATE`, `PI_KNOWLEDGE_RERANKER_API_ENDPOINT`, `PI_KNOWLEDGE_RERANKER_API_BASE_URL`, `PI_KNOWLEDGE_RERANKER_API_KEY`, `PI_KNOWLEDGE_RERANKER_API_FORMAT`, `PI_KNOWLEDGE_RERANKER_API_TIMEOUT_MS`, `PI_KNOWLEDGE_RERANKER_MAX_DOC_CHARS`, `PI_KNOWLEDGE_RERANKER_API_RESULTS_PATH`, `PI_KNOWLEDGE_RERANKER_API_INDEX_FIELD`, `PI_KNOWLEDGE_RERANKER_API_SCORE_FIELD`, `PI_KNOWLEDGE_RERANKER_API_SCORE_DIRECTION` |
|
|
182
|
+
| Reranker provider | `PI_KNOWLEDGE_RERANKER`, `PI_KNOWLEDGE_RERANKER_REVISION`, `PI_KNOWLEDGE_RERANKER_DTYPE`, `PI_KNOWLEDGE_RERANKER_REMOTE_HOST`, `PI_KNOWLEDGE_RERANKER_REMOTE_PATH_TEMPLATE`, `PI_KNOWLEDGE_RERANKER_RAW_LOGITS`, `PI_KNOWLEDGE_RERANKER_API_ENDPOINT`, `PI_KNOWLEDGE_RERANKER_API_BASE_URL`, `PI_KNOWLEDGE_RERANKER_API_KEY`, `PI_KNOWLEDGE_RERANKER_API_FORMAT`, `PI_KNOWLEDGE_RERANKER_API_TIMEOUT_MS`, `PI_KNOWLEDGE_RERANKER_MAX_DOC_CHARS`, `PI_KNOWLEDGE_RERANKER_API_RESULTS_PATH`, `PI_KNOWLEDGE_RERANKER_API_INDEX_FIELD`, `PI_KNOWLEDGE_RERANKER_API_SCORE_FIELD`, `PI_KNOWLEDGE_RERANKER_API_SCORE_DIRECTION` |
|
|
183
183
|
| Native lifecycle | `PI_KNOWLEDGE_ENABLE_NATIVE_IDLE_DISPOSE`, `PI_KNOWLEDGE_EMBEDDING_IDLE_MS` |
|
|
184
184
|
| Runtime features | `PI_KNOWLEDGE_WATCH`, `PI_KNOWLEDGE_AUTO_INJECT`, `PI_KNOWLEDGE_STALE_INDEXING_MS`, `PI_KNOWLEDGE_OFFLINE`, `PI_KNOWLEDGE_SEARCH_PROFILE`, `PI_KNOWLEDGE_SEARCH_DEFAULT_LIMIT`, `PI_KNOWLEDGE_SNIPPET_MAX_LENGTH`, `PI_KNOWLEDGE_MIN_HYBRID_SCORE`, `PI_KNOWLEDGE_SEARCH_CANDIDATE_MIN`, `PI_KNOWLEDGE_SEARCH_CANDIDATE_MULTIPLIER`, `PI_KNOWLEDGE_ADAPTIVE_CONTEXT_LINES`, `PI_KNOWLEDGE_ADAPTIVE_MAX_CHARS`, `PI_KNOWLEDGE_ADAPTIVE_NEIGHBOR_TARGET`, `PI_KNOWLEDGE_DEEP_RERANK_CANDIDATES`, `PI_KNOWLEDGE_DEEP_RERANK_TOPK_MULTIPLIER` |
|
|
185
185
|
| Release fixtures | `PI_KNOWLEDGE_E2E_PDF`, `PI_KNOWLEDGE_E2E_DOCX` |
|
|
186
186
|
|
|
187
|
+
For Hugging Face Text Embeddings Inference rerankers, use `PI_KNOWLEDGE_RERANKER=api:<model>`, `PI_KNOWLEDGE_RERANKER_API_FORMAT=tei`, and `PI_KNOWLEDGE_RERANKER_API_ENDPOINT=http://127.0.0.1:8080/rerank`; TEI exposes OpenAI-compatible embeddings under `/v1`, but reranking uses `/rerank`.
|
|
188
|
+
|
|
187
189
|
## Pi and OMP Support
|
|
188
190
|
|
|
189
191
|
`pi-knowledge` supports Pi and [OMP](https://omp.sh/) extension loading through the packaged `extension.js` entry shim. The entry stays startup-light: install-time validation can inspect the extension without resolving native runtime dependencies, and runtime modules load lazily only when tools or lifecycle hooks need them.
|
package/dist/src/model-worker.js
CHANGED
|
@@ -1,6 +1,7 @@
|
|
|
1
1
|
import { join } from "node:path";
|
|
2
2
|
import { createInterface } from "node:readline";
|
|
3
3
|
import { DEFAULT_RERANKER_REMOTE_HOST, DEFAULT_RERANKER_REMOTE_PATH_TEMPLATE, isHfRerankerConfig, rerankerCacheKey, resolveRerankerConfig, } from "./search/reranker-config.js";
|
|
4
|
+
import { rerankWithSingleRawLogit } from "./search/reranker-logits.js";
|
|
4
5
|
import { getDefaultKnowledgeDir } from "./storage/sqlite.js";
|
|
5
6
|
let embeddingPipeline = null;
|
|
6
7
|
let rerankerPipeline = null;
|
|
@@ -37,6 +38,21 @@ async function loadRerankerPipeline(config) {
|
|
|
37
38
|
const key = rerankerCacheKey(config);
|
|
38
39
|
if (rerankerPipeline?.key === key)
|
|
39
40
|
return rerankerPipeline.pipe;
|
|
41
|
+
if (config.rawLogits) {
|
|
42
|
+
const { AutoTokenizer, AutoModelForSequenceClassification, env } = await import("@huggingface/transformers");
|
|
43
|
+
const transformersEnv = env;
|
|
44
|
+
configureTransformersEnv(transformersEnv);
|
|
45
|
+
transformersEnv.remoteHost = config.remoteHost ?? DEFAULT_RERANKER_REMOTE_HOST;
|
|
46
|
+
transformersEnv.remotePathTemplate = config.remotePathTemplate ?? DEFAULT_RERANKER_REMOTE_PATH_TEMPLATE;
|
|
47
|
+
const loadOpts = { revision: config.revision };
|
|
48
|
+
if (config.dtype)
|
|
49
|
+
loadOpts.dtype = config.dtype;
|
|
50
|
+
const tokenizer = (await AutoTokenizer.from_pretrained(config.model, loadOpts));
|
|
51
|
+
const model = (await AutoModelForSequenceClassification.from_pretrained(config.model, loadOpts));
|
|
52
|
+
const pipe = (input) => rerankWithSingleRawLogit(input, tokenizer, model, config.model);
|
|
53
|
+
rerankerPipeline = { key, pipe };
|
|
54
|
+
return pipe;
|
|
55
|
+
}
|
|
40
56
|
const { pipeline, env } = await import("@huggingface/transformers");
|
|
41
57
|
const transformersEnv = env;
|
|
42
58
|
configureTransformersEnv(transformersEnv);
|
|
@@ -37,12 +37,12 @@ function mapApiResult(result, candidates, config) {
|
|
|
37
37
|
return { chunkId: candidates[index].chunkId, score };
|
|
38
38
|
}
|
|
39
39
|
function requestBody(query, candidates, topK, config) {
|
|
40
|
-
const
|
|
41
|
-
|
|
42
|
-
|
|
43
|
-
|
|
44
|
-
return
|
|
45
|
-
});
|
|
40
|
+
const texts = candidates.map((candidate) => candidate.content.length > config.maxDocumentChars
|
|
41
|
+
? candidate.content.slice(0, config.maxDocumentChars)
|
|
42
|
+
: candidate.content);
|
|
43
|
+
if (config.format === "tei")
|
|
44
|
+
return JSON.stringify({ query, texts });
|
|
45
|
+
const documents = config.format === "custom-json" ? texts.map((text, index) => ({ id: candidates[index].chunkId, text })) : texts;
|
|
46
46
|
return JSON.stringify({
|
|
47
47
|
model: config.model,
|
|
48
48
|
query,
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
export type RerankerProvider = "hf" | "api";
|
|
2
|
-
export type ApiRerankerFormat = "cohere" | "jina" | "custom-json";
|
|
2
|
+
export type ApiRerankerFormat = "cohere" | "jina" | "custom-json" | "tei";
|
|
3
3
|
export type ApiScoreDirection = "desc" | "asc";
|
|
4
4
|
export interface HfRerankerConfig {
|
|
5
5
|
provider: "hf";
|
|
@@ -8,6 +8,9 @@ export interface HfRerankerConfig {
|
|
|
8
8
|
dtype?: string;
|
|
9
9
|
remoteHost?: string;
|
|
10
10
|
remotePathTemplate?: string;
|
|
11
|
+
/** When true, use raw logits from the model instead of pipeline sigmoid scores.
|
|
12
|
+
* Essential for cross-encoder models (e.g. bge-reranker) whose logits saturate through sigmoid. */
|
|
13
|
+
rawLogits?: boolean;
|
|
11
14
|
}
|
|
12
15
|
export interface ApiRerankerConfig {
|
|
13
16
|
provider: "api";
|
|
@@ -60,8 +60,9 @@ function normalizeApiFormat(value) {
|
|
|
60
60
|
if (value === undefined)
|
|
61
61
|
return "cohere";
|
|
62
62
|
const normalized = value.toLowerCase();
|
|
63
|
-
if (normalized === "cohere" || normalized === "jina" || normalized === "custom-json")
|
|
63
|
+
if (normalized === "cohere" || normalized === "jina" || normalized === "custom-json" || normalized === "tei") {
|
|
64
64
|
return normalized;
|
|
65
|
+
}
|
|
65
66
|
throw new Error(`Unsupported reranker API format: ${value}`);
|
|
66
67
|
}
|
|
67
68
|
function normalizeScoreDirection(value) {
|
|
@@ -87,17 +88,18 @@ function resolveApiRerankerConfig(model, env) {
|
|
|
87
88
|
endpointFromBaseUrl(cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_BASE_URL) ??
|
|
88
89
|
cleanEnv(env.OPENAI_BASE_URL) ??
|
|
89
90
|
"https://api.cohere.com/v2"));
|
|
91
|
+
const format = normalizeApiFormat(cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_FORMAT));
|
|
90
92
|
return {
|
|
91
93
|
provider: "api",
|
|
92
94
|
model,
|
|
93
95
|
endpoint,
|
|
94
96
|
apiKey: cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_KEY) ?? cleanEnv(env.OPENAI_API_KEY),
|
|
95
|
-
format
|
|
97
|
+
format,
|
|
96
98
|
timeoutMs: parseNumber(cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_TIMEOUT_MS), DEFAULT_RERANKER_API_TIMEOUT_MS, 1_000, 300_000),
|
|
97
99
|
maxDocumentChars: parseNumber(cleanEnv(env.PI_KNOWLEDGE_RERANKER_MAX_DOC_CHARS), DEFAULT_RERANKER_MAX_DOC_CHARS, 100, 200_000),
|
|
98
|
-
resultsPath: cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_RESULTS_PATH) ?? "results",
|
|
100
|
+
resultsPath: cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_RESULTS_PATH) ?? (format === "tei" ? "" : "results"),
|
|
99
101
|
indexField: cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_INDEX_FIELD) ?? "index",
|
|
100
|
-
scoreField: cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_SCORE_FIELD) ?? "relevance_score",
|
|
102
|
+
scoreField: cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_SCORE_FIELD) ?? (format === "tei" ? "score" : "relevance_score"),
|
|
101
103
|
scoreDirection: normalizeScoreDirection(cleanEnv(env.PI_KNOWLEDGE_RERANKER_API_SCORE_DIRECTION)),
|
|
102
104
|
};
|
|
103
105
|
}
|
|
@@ -117,6 +119,7 @@ export function resolveRerankerConfig(env = process.env) {
|
|
|
117
119
|
throw new Error("PI_KNOWLEDGE_RERANKER requires a model id or URL");
|
|
118
120
|
const resolved = resolveHfModel(modelSource);
|
|
119
121
|
const revision = cleanEnv(env.PI_KNOWLEDGE_RERANKER_REVISION) ?? resolved.revision ?? DEFAULT_RERANKER_REVISION;
|
|
122
|
+
const rawLogits = cleanEnv(env.PI_KNOWLEDGE_RERANKER_RAW_LOGITS) === "true";
|
|
120
123
|
return {
|
|
121
124
|
provider: "hf",
|
|
122
125
|
model: resolved.model,
|
|
@@ -124,6 +127,7 @@ export function resolveRerankerConfig(env = process.env) {
|
|
|
124
127
|
dtype: cleanEnv(env.PI_KNOWLEDGE_RERANKER_DTYPE),
|
|
125
128
|
remoteHost: cleanEnv(env.PI_KNOWLEDGE_RERANKER_REMOTE_HOST) ?? resolved.remoteHost,
|
|
126
129
|
remotePathTemplate: cleanEnv(env.PI_KNOWLEDGE_RERANKER_REMOTE_PATH_TEMPLATE),
|
|
130
|
+
...(rawLogits ? { rawLogits: true } : {}),
|
|
127
131
|
};
|
|
128
132
|
}
|
|
129
133
|
export function isHfRerankerConfig(value) {
|
|
@@ -137,7 +141,8 @@ export function isHfRerankerConfig(value) {
|
|
|
137
141
|
config.revision.length > 0 &&
|
|
138
142
|
(config.dtype === undefined || typeof config.dtype === "string") &&
|
|
139
143
|
(config.remoteHost === undefined || typeof config.remoteHost === "string") &&
|
|
140
|
-
(config.remotePathTemplate === undefined || typeof config.remotePathTemplate === "string")
|
|
144
|
+
(config.remotePathTemplate === undefined || typeof config.remotePathTemplate === "string") &&
|
|
145
|
+
(config.rawLogits === undefined || typeof config.rawLogits === "boolean"));
|
|
141
146
|
}
|
|
142
147
|
export function rerankerCacheKey(config) {
|
|
143
148
|
return [
|
|
@@ -147,5 +152,6 @@ export function rerankerCacheKey(config) {
|
|
|
147
152
|
config.dtype ?? "",
|
|
148
153
|
config.remoteHost ?? "",
|
|
149
154
|
config.remotePathTemplate ?? "",
|
|
155
|
+
config.rawLogits ? "raw" : "sigmoid",
|
|
150
156
|
].join("\u0000");
|
|
151
157
|
}
|
|
@@ -0,0 +1,20 @@
|
|
|
1
|
+
export type RawLogitRerankInput = {
|
|
2
|
+
text: string;
|
|
3
|
+
text_pair: string;
|
|
4
|
+
};
|
|
5
|
+
export type RawLogitTokenizer = (text: string, options: {
|
|
6
|
+
text_pair: string;
|
|
7
|
+
padding: true;
|
|
8
|
+
truncation: true;
|
|
9
|
+
}) => Record<string, unknown>;
|
|
10
|
+
export type RawLogitModel = (inputs: Record<string, unknown>) => Promise<{
|
|
11
|
+
logits: {
|
|
12
|
+
data: ArrayLike<number>;
|
|
13
|
+
};
|
|
14
|
+
}>;
|
|
15
|
+
export declare function scoreSingleRawLogit(logits: {
|
|
16
|
+
data: ArrayLike<number>;
|
|
17
|
+
}, model: string): number;
|
|
18
|
+
export declare function rerankWithSingleRawLogit(input: RawLogitRerankInput, tokenizer: RawLogitTokenizer, model: RawLogitModel, modelName: string): Promise<Array<{
|
|
19
|
+
score: number;
|
|
20
|
+
}>>;
|
|
@@ -0,0 +1,16 @@
|
|
|
1
|
+
export function scoreSingleRawLogit(logits, model) {
|
|
2
|
+
const scores = Array.from(logits.data);
|
|
3
|
+
if (scores.length !== 1) {
|
|
4
|
+
throw new Error(`Raw-logit reranker expected a single-logit sequence classifier for ${model}, got ${scores.length} logits. ` +
|
|
5
|
+
"Use a single-logit reranker model or disable PI_KNOWLEDGE_RERANKER_RAW_LOGITS.");
|
|
6
|
+
}
|
|
7
|
+
const score = scores[0];
|
|
8
|
+
if (!Number.isFinite(score))
|
|
9
|
+
throw new Error(`Raw-logit reranker returned a non-finite score for ${model}`);
|
|
10
|
+
return score;
|
|
11
|
+
}
|
|
12
|
+
export async function rerankWithSingleRawLogit(input, tokenizer, model, modelName) {
|
|
13
|
+
const inputs = tokenizer(input.text, { text_pair: input.text_pair, padding: true, truncation: true });
|
|
14
|
+
const { logits } = await model(inputs);
|
|
15
|
+
return [{ score: scoreSingleRawLogit(logits, modelName) }];
|
|
16
|
+
}
|
package/docs/configuration.md
CHANGED
|
@@ -41,18 +41,21 @@ Each KB stores the embedding model label, vector dimension, and a non-secret emb
|
|
|
41
41
|
| `PI_KNOWLEDGE_RERANKER_DTYPE` | unset | Optional Transformers.js dtype, such as `fp32`, for local/HF rerankers. |
|
|
42
42
|
| `PI_KNOWLEDGE_RERANKER_REMOTE_HOST` | Hugging Face host or URL-derived host | Remote artifact host for HF-compatible model repositories. Use with trusted mirrors only. |
|
|
43
43
|
| `PI_KNOWLEDGE_RERANKER_REMOTE_PATH_TEMPLATE` | `{model}/resolve/{revision}/` | Transformers.js remote path template for HF-compatible mirrors. |
|
|
44
|
+
| `PI_KNOWLEDGE_RERANKER_RAW_LOGITS` | unset | Set to `true` for single-logit cross-encoder rerankers whose raw scores are flattened by sigmoid scoring. Multi-logit classifiers are rejected unless this option is disabled. |
|
|
44
45
|
| `PI_KNOWLEDGE_RERANKER_API_ENDPOINT` | derived from base URL | Full HTTP rerank endpoint. Takes precedence over `PI_KNOWLEDGE_RERANKER_API_BASE_URL` and `OPENAI_BASE_URL`. |
|
|
45
46
|
| `PI_KNOWLEDGE_RERANKER_API_BASE_URL` | `https://api.cohere.com/v2` | Base URL for Cohere/Jina-compatible `/rerank` APIs. `OPENAI_BASE_URL` is accepted only as a convenience fallback for gateways exposing the same schema. |
|
|
46
47
|
| `PI_KNOWLEDGE_RERANKER_API_KEY` | `OPENAI_API_KEY` fallback | Bearer token for API rerankers. Omit only for trusted local services that do not require auth. |
|
|
47
|
-
| `PI_KNOWLEDGE_RERANKER_API_FORMAT` | `cohere` | Request/response preset: `cohere`, `jina`,
|
|
48
|
+
| `PI_KNOWLEDGE_RERANKER_API_FORMAT` | `cohere` | Request/response preset: `cohere`, `jina`, `custom-json`, or `tei`. Cohere/Jina-compatible presets send `model`, `query`, `documents`, `top_n`, and `return_documents:false`; `tei` sends `query` plus `texts` and parses a root response array with `index`/`score`. |
|
|
48
49
|
| `PI_KNOWLEDGE_RERANKER_API_TIMEOUT_MS` | `30000` | Per-request timeout for API reranking. |
|
|
49
50
|
| `PI_KNOWLEDGE_RERANKER_MAX_DOC_CHARS` | `12000` | Per-candidate document safety cap before sending content to the API. |
|
|
50
|
-
| `PI_KNOWLEDGE_RERANKER_API_RESULTS_PATH` | `results` | Dot path to the result array for `custom-json` or compatible responses. |
|
|
51
|
+
| `PI_KNOWLEDGE_RERANKER_API_RESULTS_PATH` | `results` (`tei`: root response array) | Dot path to the result array for `custom-json` or compatible responses. |
|
|
51
52
|
| `PI_KNOWLEDGE_RERANKER_API_INDEX_FIELD` | `index` | Field name or dot path holding the original candidate index. |
|
|
52
|
-
| `PI_KNOWLEDGE_RERANKER_API_SCORE_FIELD` | `relevance_score` | Field name or dot path holding the rerank score. |
|
|
53
|
+
| `PI_KNOWLEDGE_RERANKER_API_SCORE_FIELD` | `relevance_score` (`tei`: `score`) | Field name or dot path holding the rerank score. |
|
|
53
54
|
| `PI_KNOWLEDGE_RERANKER_API_SCORE_DIRECTION` | `desc` | Sort direction for returned scores: `desc` when higher is better, `asc` when lower is better. |
|
|
54
55
|
|
|
55
|
-
Reranker settings affect query-time `deep` search only; they do not change indexed KB vectors. Local/HF rerankers run in the isolated model worker. Offline mode sets the model worker to local-only loading, so a local/HF reranker must already exist in the model cache. API rerankers send the query and selected candidate chunk content to the configured external service; failures surface by default and do not silently fall back to the local worker.
|
|
56
|
+
Reranker settings affect query-time `deep` search only; they do not change indexed KB vectors. Local/HF rerankers run in the isolated model worker. Offline mode sets the model worker to local-only loading, so a local/HF reranker must already exist in the model cache. `PI_KNOWLEDGE_RERANKER_RAW_LOGITS=true` switches local/HF rerankers from the Transformers.js text-classification pipeline to direct sequence-classification logits and is intended only for trusted single-logit reranker models; use trusted model repositories or pinned revisions for non-default artifacts. API rerankers send the query and selected candidate chunk content to the configured external service; failures surface by default and do not silently fall back to the local worker.
|
|
57
|
+
|
|
58
|
+
For Hugging Face Text Embeddings Inference rerankers, set `PI_KNOWLEDGE_RERANKER_API_FORMAT=tei` and point `PI_KNOWLEDGE_RERANKER_API_ENDPOINT` at the TEI `/rerank` endpoint, for example `http://127.0.0.1:8080/rerank`. Do not reuse an embeddings `OPENAI_BASE_URL` ending in `/v1` unless the reranker endpoint actually exists there; TEI's OpenAI-compatible route covers embeddings, while reranking uses `/rerank`.
|
|
56
59
|
|
|
57
60
|
Local embeddings require a real Node 22+ executable for the isolated model worker. `pi-knowledge` first tries Node IPC and automatically falls back to a stdin/stdout JSONL worker transport when the host runtime does not expose `child_process.fork().send()`, which can happen in Windows OMP compatibility layers. On Windows it searches `PI_KNOWLEDGE_NODE_PATH`, persisted `knowledge_configure` settings, common Node, Volta, NVM, Codex `cua_node`, and `PATH` locations before failing. If worker startup still fails, call `knowledge_configure` with the full `node.exe` path such as `C:\Program Files\nodejs\node.exe`; accidental surrounding quotes are ignored. OpenAI-compatible embeddings avoid the local worker entirely.
|
|
58
61
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "pi-knowledge",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.8.0",
|
|
4
4
|
"description": "Local-first RAG knowledge base for Pi and OMP coding agents with hybrid code/doc search, reranking, diagnostics, and large-project indexing",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "dist/index.js",
|