pi-braid 0.43.6 → 0.43.8
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 +83 -0
- package/TOOLS.md +1 -1
- package/build-info.json +4 -4
- package/dist/extensions/core/prompt-builder/tool-guidelines.js +1 -1
- package/dist/extensions/core/skills-tool/applicability.d.ts +134 -0
- package/dist/extensions/core/skills-tool/applicability.js +161 -0
- package/dist/extensions/core/skills-tool/auto-inject.d.ts +31 -0
- package/dist/extensions/core/skills-tool/auto-inject.js +320 -1
- package/dist/extensions/core/skills-tool/index.js +83 -4
- package/dist/extensions/core/skills-tool/loading-guidance.d.ts +4 -1
- package/dist/extensions/core/skills-tool/loading-guidance.js +6 -2
- package/dist/extensions/core/skills-tool/session-records.d.ts +395 -8
- package/dist/extensions/core/skills-tool/session-records.js +645 -5
- package/dist/extensions/core/skills-tool/skill-families.d.ts +60 -0
- package/dist/extensions/core/skills-tool/skill-families.js +105 -0
- package/dist/extensions/enforcement/handlers/analysis/dreamer-handler.js +5 -1
- package/dist/extensions/enforcement/handlers/register-tool-call.js +5 -2
- package/dist/extensions/enforcement/verification-judge.d.ts +7 -0
- package/dist/extensions/enforcement/verification-judge.js +28 -21
- package/dist/extensions/memory/session-recall/ask.d.ts +95 -0
- package/dist/extensions/memory/session-recall/ask.js +376 -0
- package/dist/extensions/memory/session-recall/corpus-index.d.ts +200 -0
- package/dist/extensions/memory/session-recall/corpus-index.js +480 -0
- package/dist/extensions/memory/session-recall/index.js +173 -21
- package/dist/extensions/memory/session-recall/search.d.ts +29 -0
- package/dist/extensions/memory/session-recall/search.js +42 -9
- package/dist/lib/ai/payload-hygiene.d.ts +3 -0
- package/dist/lib/ai/payload-hygiene.js +34 -2
- package/dist/lib/config/braid-paths.d.ts +4 -0
- package/dist/lib/config/braid-paths.js +6 -0
- package/dist/lib/config/braid-settings.d.ts +22 -0
- package/dist/lib/config/braid-settings.js +15 -0
- package/dist/lib/detection/tool-loop-detector.d.ts +95 -46
- package/dist/lib/detection/tool-loop-detector.js +288 -180
- package/dist/lib/reasoning/dreamer.d.ts +2 -1
- package/dist/lib/reasoning/dreamer.js +12 -7
- package/docs/LOCAL_DEV.md +1 -1
- package/docs/SETUP.md +1 -0
- package/package.json +1 -1
- package/src/extensions/core/AGENTS.md +3 -1
- package/src/extensions/core/prompt-builder/tool-guidelines.ts +1 -1
- package/src/extensions/core/skills-tool/applicability.ts +220 -0
- package/src/extensions/core/skills-tool/auto-inject.ts +391 -0
- package/src/extensions/core/skills-tool/index.ts +96 -3
- package/src/extensions/core/skills-tool/loading-guidance.ts +6 -2
- package/src/extensions/core/skills-tool/session-records.ts +932 -5
- package/src/extensions/core/skills-tool/skill-families.ts +112 -0
- package/src/extensions/enforcement/handlers/analysis/dreamer-handler.ts +5 -1
- package/src/extensions/enforcement/handlers/register-tool-call.ts +5 -2
- package/src/extensions/enforcement/verification-judge.ts +32 -25
- package/src/extensions/memory/AGENTS.md +2 -2
- package/src/extensions/memory/session-recall/AGENTS.md +92 -18
- package/src/extensions/memory/session-recall/ask.ts +435 -0
- package/src/extensions/memory/session-recall/corpus-index.ts +609 -0
- package/src/extensions/memory/session-recall/index.ts +202 -21
- package/src/extensions/memory/session-recall/search.ts +61 -10
- package/src/lib/ai/payload-hygiene.ts +47 -3
- package/src/lib/config/braid-paths.ts +7 -0
- package/src/lib/config/braid-settings.ts +42 -3
- package/src/lib/detection/AGENTS.md +8 -7
- package/src/lib/detection/tool-loop-detector.ts +339 -195
- package/src/lib/reasoning/dreamer.ts +13 -6
|
@@ -0,0 +1,60 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* The tool families a skill declares it governs.
|
|
3
|
+
*
|
|
4
|
+
* A trigger phrase says WHEN to open a skill file; nothing in the file says
|
|
5
|
+
* what the file's advice is then used FOR. `governs:` closes that: the tool
|
|
6
|
+
* families whose calls the body is written to direct. A family is the tool
|
|
7
|
+
* name's first underscore-delimited segment (`search_documents` → `search`),
|
|
8
|
+
* the same token `loading-guidance.ts` pairs with a skill name in the
|
|
9
|
+
* mechanical trigger, so "before the first call in the family" in the prompt
|
|
10
|
+
* and "a call in a declared family" in the record cannot mean two things.
|
|
11
|
+
*
|
|
12
|
+
* The declaration is authored data in the skill's own frontmatter — pi's
|
|
13
|
+
* parser already returns it, and the pack's authoring contract already admits
|
|
14
|
+
* any YAML-valid extra key. No new file format and no new field parser is
|
|
15
|
+
* needed, which is why this module only reads.
|
|
16
|
+
*
|
|
17
|
+
* Most skills govern nothing: advice about prose, a palette or a checklist
|
|
18
|
+
* directs no tool, and a file without the key is read as declaring no family.
|
|
19
|
+
* The absence is a fact, not a gap — an obligation that names no family is
|
|
20
|
+
* satisfied by the read alone.
|
|
21
|
+
*
|
|
22
|
+
* A declared family may not be the detector's own tool (`read`): a call to it
|
|
23
|
+
* IS the read leg, so counting it as an applied call would let an obligation
|
|
24
|
+
* satisfy itself. The parser drops such an entry; the shipped pack's contract
|
|
25
|
+
* refuses it by name so an author sees the problem before a release.
|
|
26
|
+
*
|
|
27
|
+
* Nothing here names a skill, a package or a path. Reads are memoized by file
|
|
28
|
+
* mtime + size, so a turn pays one stat per skill file and a parse only when
|
|
29
|
+
* a file changed.
|
|
30
|
+
*/
|
|
31
|
+
/**
|
|
32
|
+
* Max families read from one declaration. A body that directs more than four
|
|
33
|
+
* tool families is not one skill's job, and the record's per-session rows are
|
|
34
|
+
* bounded by the same number. The pack's contract mirrors this cap.
|
|
35
|
+
*/
|
|
36
|
+
export declare const MAX_GOVERNED_FAMILIES = 4;
|
|
37
|
+
/** The token shape a declaration must use: a tool name's first segment. */
|
|
38
|
+
export declare const GOVERNED_FAMILY_SHAPE: RegExp;
|
|
39
|
+
/**
|
|
40
|
+
* The tool whose call is the read leg itself. Never a declarable family: see
|
|
41
|
+
* the module header.
|
|
42
|
+
*/
|
|
43
|
+
export declare const DETECTOR_TOOL = "read";
|
|
44
|
+
/** The family a tool name belongs to: its first underscore-delimited segment. */
|
|
45
|
+
export declare function toolFamilyToken(toolName: string): string;
|
|
46
|
+
/**
|
|
47
|
+
* Normalize a frontmatter `governs` value into family tokens. Accepts a YAML
|
|
48
|
+
* list or a comma-separated string, lowercases, drops malformed, detector, and
|
|
49
|
+
* surplus entries, and dedupes in authored order. Pure.
|
|
50
|
+
*/
|
|
51
|
+
export declare function parseGovernedFamilies(value: unknown): string[];
|
|
52
|
+
/**
|
|
53
|
+
* Read the families a skill file declares. Missing or unreadable file, or
|
|
54
|
+
* malformed frontmatter, yields an empty list (a declaration that cannot be
|
|
55
|
+
* read is not a declaration). Memoized: an unchanged file never re-reads.
|
|
56
|
+
*/
|
|
57
|
+
export declare function readGovernedFamilies(filePath: string): string[];
|
|
58
|
+
/** Test seam: forget every memoized read. */
|
|
59
|
+
export declare function clearGovernedFamilyCache(): void;
|
|
60
|
+
//# sourceMappingURL=skill-families.d.ts.map
|
|
@@ -0,0 +1,105 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* The tool families a skill declares it governs.
|
|
3
|
+
*
|
|
4
|
+
* A trigger phrase says WHEN to open a skill file; nothing in the file says
|
|
5
|
+
* what the file's advice is then used FOR. `governs:` closes that: the tool
|
|
6
|
+
* families whose calls the body is written to direct. A family is the tool
|
|
7
|
+
* name's first underscore-delimited segment (`search_documents` → `search`),
|
|
8
|
+
* the same token `loading-guidance.ts` pairs with a skill name in the
|
|
9
|
+
* mechanical trigger, so "before the first call in the family" in the prompt
|
|
10
|
+
* and "a call in a declared family" in the record cannot mean two things.
|
|
11
|
+
*
|
|
12
|
+
* The declaration is authored data in the skill's own frontmatter — pi's
|
|
13
|
+
* parser already returns it, and the pack's authoring contract already admits
|
|
14
|
+
* any YAML-valid extra key. No new file format and no new field parser is
|
|
15
|
+
* needed, which is why this module only reads.
|
|
16
|
+
*
|
|
17
|
+
* Most skills govern nothing: advice about prose, a palette or a checklist
|
|
18
|
+
* directs no tool, and a file without the key is read as declaring no family.
|
|
19
|
+
* The absence is a fact, not a gap — an obligation that names no family is
|
|
20
|
+
* satisfied by the read alone.
|
|
21
|
+
*
|
|
22
|
+
* A declared family may not be the detector's own tool (`read`): a call to it
|
|
23
|
+
* IS the read leg, so counting it as an applied call would let an obligation
|
|
24
|
+
* satisfy itself. The parser drops such an entry; the shipped pack's contract
|
|
25
|
+
* refuses it by name so an author sees the problem before a release.
|
|
26
|
+
*
|
|
27
|
+
* Nothing here names a skill, a package or a path. Reads are memoized by file
|
|
28
|
+
* mtime + size, so a turn pays one stat per skill file and a parse only when
|
|
29
|
+
* a file changed.
|
|
30
|
+
*/
|
|
31
|
+
import { readFileSync, statSync } from "node:fs";
|
|
32
|
+
import { parseFrontmatter } from "@earendil-works/pi-coding-agent";
|
|
33
|
+
/**
|
|
34
|
+
* Max families read from one declaration. A body that directs more than four
|
|
35
|
+
* tool families is not one skill's job, and the record's per-session rows are
|
|
36
|
+
* bounded by the same number. The pack's contract mirrors this cap.
|
|
37
|
+
*/
|
|
38
|
+
export const MAX_GOVERNED_FAMILIES = 4;
|
|
39
|
+
/** The token shape a declaration must use: a tool name's first segment. */
|
|
40
|
+
export const GOVERNED_FAMILY_SHAPE = /^[a-z][a-z0-9]*$/;
|
|
41
|
+
/**
|
|
42
|
+
* The tool whose call is the read leg itself. Never a declarable family: see
|
|
43
|
+
* the module header.
|
|
44
|
+
*/
|
|
45
|
+
export const DETECTOR_TOOL = "read";
|
|
46
|
+
/** The family a tool name belongs to: its first underscore-delimited segment. */
|
|
47
|
+
export function toolFamilyToken(toolName) {
|
|
48
|
+
return (toolName ?? "").split("_")[0].toLowerCase();
|
|
49
|
+
}
|
|
50
|
+
/**
|
|
51
|
+
* Normalize a frontmatter `governs` value into family tokens. Accepts a YAML
|
|
52
|
+
* list or a comma-separated string, lowercases, drops malformed, detector, and
|
|
53
|
+
* surplus entries, and dedupes in authored order. Pure.
|
|
54
|
+
*/
|
|
55
|
+
export function parseGovernedFamilies(value) {
|
|
56
|
+
const raw = Array.isArray(value) ? value : typeof value === "string" ? value.split(",") : [];
|
|
57
|
+
const families = [];
|
|
58
|
+
for (const item of raw) {
|
|
59
|
+
if (typeof item !== "string")
|
|
60
|
+
continue;
|
|
61
|
+
const family = item.trim().toLowerCase();
|
|
62
|
+
if (!GOVERNED_FAMILY_SHAPE.test(family) || family === DETECTOR_TOOL)
|
|
63
|
+
continue;
|
|
64
|
+
if (families.includes(family))
|
|
65
|
+
continue;
|
|
66
|
+
families.push(family);
|
|
67
|
+
if (families.length >= MAX_GOVERNED_FAMILIES)
|
|
68
|
+
break;
|
|
69
|
+
}
|
|
70
|
+
return families;
|
|
71
|
+
}
|
|
72
|
+
const cache = new Map();
|
|
73
|
+
/**
|
|
74
|
+
* Read the families a skill file declares. Missing or unreadable file, or
|
|
75
|
+
* malformed frontmatter, yields an empty list (a declaration that cannot be
|
|
76
|
+
* read is not a declaration). Memoized: an unchanged file never re-reads.
|
|
77
|
+
*/
|
|
78
|
+
export function readGovernedFamilies(filePath) {
|
|
79
|
+
let st;
|
|
80
|
+
try {
|
|
81
|
+
st = statSync(filePath);
|
|
82
|
+
}
|
|
83
|
+
catch {
|
|
84
|
+
cache.delete(filePath);
|
|
85
|
+
return [];
|
|
86
|
+
}
|
|
87
|
+
const cached = cache.get(filePath);
|
|
88
|
+
if (cached && cached.mtimeMs === st.mtimeMs && cached.size === st.size)
|
|
89
|
+
return cached.families;
|
|
90
|
+
let families = [];
|
|
91
|
+
try {
|
|
92
|
+
const { frontmatter } = parseFrontmatter(readFileSync(filePath, "utf-8"));
|
|
93
|
+
families = parseGovernedFamilies(frontmatter.governs);
|
|
94
|
+
}
|
|
95
|
+
catch {
|
|
96
|
+
// Unreadable or unparseable: no declared family, and no throw on a turn.
|
|
97
|
+
}
|
|
98
|
+
cache.set(filePath, { mtimeMs: st.mtimeMs, size: st.size, families });
|
|
99
|
+
return families;
|
|
100
|
+
}
|
|
101
|
+
/** Test seam: forget every memoized read. */
|
|
102
|
+
export function clearGovernedFamilyCache() {
|
|
103
|
+
cache.clear();
|
|
104
|
+
}
|
|
105
|
+
//# sourceMappingURL=skill-families.js.map
|
|
@@ -7,6 +7,7 @@ import { emitSignal } from "../../../../lib/reasoning/belief-policy-engine.js";
|
|
|
7
7
|
import { dream } from "../../../../lib/reasoning/dreamer.js";
|
|
8
8
|
import { recordAdvice } from "../../../../lib/reasoning/advice-tracker.js";
|
|
9
9
|
import { tryClaim } from "../../../../lib/enforcement/turn-dedup.js";
|
|
10
|
+
import { getToolLoopState } from "../../../../lib/detection/tool-loop-detector.js";
|
|
10
11
|
import { getState } from "../../state.js";
|
|
11
12
|
const log = createLogger("enforcement:dreamer");
|
|
12
13
|
export async function dreamerHandler(pi, _event, ctx) {
|
|
@@ -20,7 +21,10 @@ export async function dreamerHandler(pi, _event, ctx) {
|
|
|
20
21
|
historicalSequences.push(turn.actualTools);
|
|
21
22
|
}
|
|
22
23
|
}
|
|
23
|
-
|
|
24
|
+
// The loop path reads the detector's own recorded calls (tool + argument
|
|
25
|
+
// identity); gs.toolLog is names only and cannot distinguish one call
|
|
26
|
+
// repeated from many different calls under one name.
|
|
27
|
+
const dreamAdvice = dream(ctx, gs.toolLog, historicalSequences, getToolLoopState(ctx).history);
|
|
24
28
|
if (!dreamAdvice)
|
|
25
29
|
return;
|
|
26
30
|
emitSignal(ctx, "stability", "predicted-decline", "info", false, {
|
|
@@ -29,9 +29,12 @@ export function registerToolCallHandler(pi) {
|
|
|
29
29
|
if (!toolName || !toolName.trim())
|
|
30
30
|
return undefined;
|
|
31
31
|
recordToolCall(toolName, ctx);
|
|
32
|
-
// Record tool call for tool-loop-detector (per-turn tool sets)
|
|
32
|
+
// Record tool call for tool-loop-detector (per-turn tool sets). The
|
|
33
|
+
// arguments ride along: a loop is the same call repeating, and the tool
|
|
34
|
+
// name alone cannot say whether sixteen `bash` calls are sixteen commands
|
|
35
|
+
// or one command sixteen times.
|
|
33
36
|
const loopState = getToolLoopState(ctx);
|
|
34
|
-
recordLoopToolCall(loopState, toolName);
|
|
37
|
+
recordLoopToolCall(loopState, toolName, input);
|
|
35
38
|
recordToolInput(toolName, input, getState(ctx));
|
|
36
39
|
if (toolName === "bash") {
|
|
37
40
|
const cmd = input?.command || "";
|
|
@@ -79,6 +79,13 @@ export interface JudgeContext {
|
|
|
79
79
|
}
|
|
80
80
|
/**
|
|
81
81
|
* Build the verification analysis prompt. Exported for tests.
|
|
82
|
+
*
|
|
83
|
+
* Order is load-bearing: the ground-truth sections (git history, the
|
|
84
|
+
* disk-checked deliverables manifest, the code-changes flag) come BEFORE the
|
|
85
|
+
* bounded ones (task evidence, tool log, reasoning, response). The bounded
|
|
86
|
+
* sections are the ones a bound can shorten: a head kept under a size cap, a
|
|
87
|
+
* tail cut by a redaction. A reader that meets the verdict-bearing sections
|
|
88
|
+
* first cannot have them removed by whatever truncated the evidence.
|
|
82
89
|
*/
|
|
83
90
|
export declare function buildVerificationPrompt(context: JudgeContext): string;
|
|
84
91
|
/**
|
|
@@ -61,6 +61,13 @@ import { createLogger } from "../../lib/core/logger.js";
|
|
|
61
61
|
const log = createLogger("verification-judge");
|
|
62
62
|
/**
|
|
63
63
|
* Build the verification analysis prompt. Exported for tests.
|
|
64
|
+
*
|
|
65
|
+
* Order is load-bearing: the ground-truth sections (git history, the
|
|
66
|
+
* disk-checked deliverables manifest, the code-changes flag) come BEFORE the
|
|
67
|
+
* bounded ones (task evidence, tool log, reasoning, response). The bounded
|
|
68
|
+
* sections are the ones a bound can shorten: a head kept under a size cap, a
|
|
69
|
+
* tail cut by a redaction. A reader that meets the verdict-bearing sections
|
|
70
|
+
* first cannot have them removed by whatever truncated the evidence.
|
|
64
71
|
*/
|
|
65
72
|
export function buildVerificationPrompt(context) {
|
|
66
73
|
const sections = [];
|
|
@@ -70,27 +77,6 @@ export function buildVerificationPrompt(context) {
|
|
|
70
77
|
sections.push("## TOOLS USED");
|
|
71
78
|
sections.push(context.toolsUsed.length > 0 ? context.toolsUsed.join(", ") : "(none)");
|
|
72
79
|
sections.push("");
|
|
73
|
-
// Task-wide evidence (since last user message, session-store
|
|
74
|
-
// backed) comes BEFORE the current-window tool log — it is the surviving
|
|
75
|
-
// view when turn breaks cut toolLog down to the tail turn.
|
|
76
|
-
if (context.taskEvidence && context.taskEvidence.length > 0) {
|
|
77
|
-
sections.push("## TASK EVIDENCE (since last user message, disk/session-surviving)");
|
|
78
|
-
sections.push(context.taskEvidence.join("\n"));
|
|
79
|
-
sections.push("");
|
|
80
|
-
}
|
|
81
|
-
if (context.toolLog.length > 0) {
|
|
82
|
-
sections.push("## TOOL EXECUTION LOG");
|
|
83
|
-
sections.push(context.toolLog.slice(-50).join("\n")); // Last 50 entries (matches turnHistory cap)
|
|
84
|
-
sections.push("");
|
|
85
|
-
}
|
|
86
|
-
sections.push(`## CODE CHANGES: ${context.hasCodeChanges ? "YES" : "NO"}`);
|
|
87
|
-
// (Verification evidence comes from the TOOL EXECUTION LOG above, not a boolean)
|
|
88
|
-
sections.push("");
|
|
89
|
-
if (context.thinking) {
|
|
90
|
-
sections.push("## AGENT REASONING (thinking blocks)");
|
|
91
|
-
sections.push(context.thinking);
|
|
92
|
-
sections.push("");
|
|
93
|
-
}
|
|
94
80
|
if (context.gitHistory) {
|
|
95
81
|
sections.push("## GIT HISTORY (recent commits)");
|
|
96
82
|
sections.push(context.gitHistory);
|
|
@@ -107,6 +93,27 @@ export function buildVerificationPrompt(context) {
|
|
|
107
93
|
: "\nSome deliverables missing on disk → real unverified risk.");
|
|
108
94
|
sections.push("");
|
|
109
95
|
}
|
|
96
|
+
sections.push(`## CODE CHANGES: ${context.hasCodeChanges ? "YES" : "NO"}`);
|
|
97
|
+
// (Verification evidence comes from the TOOL EXECUTION LOG, not a boolean)
|
|
98
|
+
sections.push("");
|
|
99
|
+
// Task-wide evidence (since last user message, session-store backed) comes
|
|
100
|
+
// BEFORE the current-window tool log — it is the surviving view when turn
|
|
101
|
+
// breaks cut toolLog down to the tail turn.
|
|
102
|
+
if (context.taskEvidence && context.taskEvidence.length > 0) {
|
|
103
|
+
sections.push("## TASK EVIDENCE (since last user message, disk/session-surviving)");
|
|
104
|
+
sections.push(context.taskEvidence.join("\n"));
|
|
105
|
+
sections.push("");
|
|
106
|
+
}
|
|
107
|
+
if (context.toolLog.length > 0) {
|
|
108
|
+
sections.push("## TOOL EXECUTION LOG");
|
|
109
|
+
sections.push(context.toolLog.slice(-50).join("\n")); // Last 50 entries (matches turnHistory cap)
|
|
110
|
+
sections.push("");
|
|
111
|
+
}
|
|
112
|
+
if (context.thinking) {
|
|
113
|
+
sections.push("## AGENT REASONING (thinking blocks)");
|
|
114
|
+
sections.push(context.thinking);
|
|
115
|
+
sections.push("");
|
|
116
|
+
}
|
|
110
117
|
if (context.lastResponse) {
|
|
111
118
|
sections.push("## AGENT RESPONSE (visible text)");
|
|
112
119
|
sections.push(context.lastResponse);
|
|
@@ -0,0 +1,95 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Session corpus ask - lexical retrieval over the user-turn index.
|
|
3
|
+
*
|
|
4
|
+
* BM25 over one document per user turn. At this corpus size the lexical model
|
|
5
|
+
* is the cheap, testable choice: no embedding pass to keep current, no vector
|
|
6
|
+
* store to back up, no key to spend on a query, and a score a reader can
|
|
7
|
+
* re-derive by eye from the terms it names. What would earn a vector store is
|
|
8
|
+
* a measured failure: recall@3 below the bar on questions whose answering
|
|
9
|
+
* passage is in the index (the report's falsifier is that measurement).
|
|
10
|
+
*
|
|
11
|
+
* Every hit carries its citation - project, session path, line, timestamp -
|
|
12
|
+
* because the answer is the owner's own words and a reader must be able to
|
|
13
|
+
* check them against the session file. The refusal is a first-class result: a
|
|
14
|
+
* query whose terms appear nowhere in the population says so and names what it
|
|
15
|
+
* searched, rather than returning the least bad rows.
|
|
16
|
+
*/
|
|
17
|
+
import type { CorpusFreshness, CorpusTurn, LoadedShard } from "./corpus-index.js";
|
|
18
|
+
/** BM25 term-frequency saturation. */
|
|
19
|
+
export declare const BM25_K1 = 1.2;
|
|
20
|
+
/** BM25 length normalisation. */
|
|
21
|
+
export declare const BM25_B = 0.75;
|
|
22
|
+
/** Hits a query returns when the caller names no limit. */
|
|
23
|
+
export declare const MAX_ASK_HITS = 5;
|
|
24
|
+
/**
|
|
25
|
+
* Words a document is scored on: letter/number runs, lowercased, stemmed.
|
|
26
|
+
*
|
|
27
|
+
* A hyphen separates words: the owner's compound names ("looping-clanker",
|
|
28
|
+
* "anti-crowd") are two words a question may spell either way, and a tokenizer
|
|
29
|
+
* that keeps the compound as one token can never match its own parts.
|
|
30
|
+
*/
|
|
31
|
+
export declare function tokenizeForIndex(text: string): string[];
|
|
32
|
+
/**
|
|
33
|
+
* Query terms: the document tokens minus function words. Duplicates collapse,
|
|
34
|
+
* so a word typed twice scores once.
|
|
35
|
+
*/
|
|
36
|
+
export declare function askQueryTerms(query: string): string[];
|
|
37
|
+
export interface AskHit {
|
|
38
|
+
turn: CorpusTurn;
|
|
39
|
+
score: number;
|
|
40
|
+
/** Query terms this turn carries. */
|
|
41
|
+
matched: string[];
|
|
42
|
+
}
|
|
43
|
+
export interface AskResult {
|
|
44
|
+
query: string;
|
|
45
|
+
/** Query terms after the function-word filter. */
|
|
46
|
+
terms: string[];
|
|
47
|
+
/** Terms that occur in the searched population. */
|
|
48
|
+
matched_terms: string[];
|
|
49
|
+
/** Terms that occur nowhere in the searched population. */
|
|
50
|
+
unmatched_terms: string[];
|
|
51
|
+
hits: AskHit[];
|
|
52
|
+
/** Whether the index could answer at all (at least one hit). */
|
|
53
|
+
answered: boolean;
|
|
54
|
+
population: {
|
|
55
|
+
turns: number;
|
|
56
|
+
sessions: number;
|
|
57
|
+
projects: string[];
|
|
58
|
+
/** Oldest and newest shard build time in the population. */
|
|
59
|
+
built_oldest: string;
|
|
60
|
+
built_newest: string;
|
|
61
|
+
shards: number;
|
|
62
|
+
};
|
|
63
|
+
freshness: CorpusFreshness;
|
|
64
|
+
filters: {
|
|
65
|
+
project: string | null;
|
|
66
|
+
since: string | null;
|
|
67
|
+
};
|
|
68
|
+
/** True when the caller's own session was withheld from the results. */
|
|
69
|
+
self_excluded: boolean;
|
|
70
|
+
}
|
|
71
|
+
export interface AskOptions {
|
|
72
|
+
/** Project label filter; null reads every shard the store loaded. */
|
|
73
|
+
project?: string | null;
|
|
74
|
+
/** Inclusive lower bound on the turn's date (YYYY-MM-DD). */
|
|
75
|
+
since?: string | null;
|
|
76
|
+
limit?: number;
|
|
77
|
+
/** Session file never returned to itself. */
|
|
78
|
+
excludeSession?: string | null;
|
|
79
|
+
freshness: CorpusFreshness;
|
|
80
|
+
}
|
|
81
|
+
/**
|
|
82
|
+
* Retrieve the turns that answer a question, with citations.
|
|
83
|
+
*
|
|
84
|
+
* The population searched is stated in the result, so a refusal can show what
|
|
85
|
+
* it looked at. `df` and the average document length are computed over the
|
|
86
|
+
* filtered population: an index narrowed by `--project` or `--since` is the
|
|
87
|
+
* corpus the answer claims, not a sample of a larger one it never read.
|
|
88
|
+
*/
|
|
89
|
+
export declare function askCorpusIndex(shards: LoadedShard[], query: string, opts: AskOptions): AskResult;
|
|
90
|
+
/**
|
|
91
|
+
* Render the answer: cited passages, or a refusal that names the population it
|
|
92
|
+
* searched and the terms it found.
|
|
93
|
+
*/
|
|
94
|
+
export declare function formatAskResult(result: AskResult): string;
|
|
95
|
+
//# sourceMappingURL=ask.d.ts.map
|