sensemaking 0.15.3 → 0.16.0

package/dist/esm/cli/download.js

@@ -1,11 +1,11 @@
+import { basename } from 'node:path';
 import { embedConfig } from '../config/index.js';
+import { downloadModel, isDownloadable, modelDir, modelPresent } from '../embed/store.js';
 import { SenseError } from '../errors.js';
-import { downloadModel, modelDir, modelPresent } from '../features/embed.js';
 import { USAGE } from './index.js';
 import { CONFIG, parse } from './shared.js';
-// The one command that fetches weights. Nothing else touches the network for them, so a
-// query that reads like a query never spends minutes pulling 124 MB. Asks nothing, so there
-// is no prompt for CI to bypass; running it twice is a no-op.
+// A prefetch over the store's own fetch path: a search would fetch the same weights
+// lazily, so this only moves the wait earlier. Idempotent; asks nothing, so CI can run it unattended.
 const download = async (ctx)=>{
     const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.download}`, {
         ...CONFIG
@@ -13,15 +13,16 @@ const download = async (ctx)=>{
     const cfg = ctx.resolveConfig(values.config);
     const e = embedConfig(cfg);
     if (!e) throw new SenseError('EMBED_DISABLED', 'no preset asks for vectors, so there is no model to download');
-    if (e.type === 'api') {
-        console.log(`embed type is "api" (${e.url}), so there is nothing to download; ${ctx.name} search reaches it over the network`);
+    if (e.provider !== 'static') {
+        console.log(`embed provider is "${e.provider}" (${e.url}), so there is nothing to download; ${ctx.name} search reaches it over the network`);
         return;
     }
     if (modelPresent(cfg)) {
-        console.log(`${e.model} is already available in ${modelDir(e.model)}`);
+        const at = isDownloadable(e.model) ? `@${basename(modelDir(e.model))} ` : ' ';
+        console.log(`${e.model}${at}is already available in ${modelDir(e.model)}`);
         return;
     }
     const dir = await downloadModel(e.model, (file, into)=>console.error(`fetching ${e.model}/${file} into ${into}`));
-    console.log(`downloaded ${e.model} into ${dir}`);
+    console.log(`downloaded ${e.model}@${basename(dir)} into ${dir}`);
 };
 export default download;

package/dist/esm/cli/download.js.map

@@ -1 +1 @@
-{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/download.ts"],"sourcesContent":["import { embedConfig } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { downloadModel, modelDir, modelPresent } from '../features/embed.ts';\nimport { USAGE } from './index.ts';\nimport { CONFIG, parse } from './shared.ts';\nimport type { Command } from './types.ts';\n\n// The one command that fetches weights. Nothing else touches the network for them, so a\n// query that reads like a query never spends minutes pulling 124 MB. Asks nothing, so there\n// is no prompt for CI to bypass; running it twice is a no-op.\nconst download: Command = async (ctx) => {\n  const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.download}`, { ...CONFIG });\n  const cfg = ctx.resolveConfig(values.config as string | undefined);\n  const e = embedConfig(cfg);\n  if (!e) throw new SenseError('EMBED_DISABLED', 'no preset asks for vectors, so there is no model to download');\n\n  if (e.type === 'api') {\n    console.log(`embed type is \"api\" (${e.url}), so there is nothing to download; ${ctx.name} search reaches it over the network`);\n    return;\n  }\n  if (modelPresent(cfg)) {\n    console.log(`${e.model} is already available in ${modelDir(e.model)}`);\n    return;\n  }\n  const dir = await downloadModel(e.model, (file, into) => console.error(`fetching ${e.model}/${file} into ${into}`));\n  console.log(`downloaded ${e.model} into ${dir}`);\n};\nexport default download;\n"],"names":["embedConfig","SenseError","downloadModel","modelDir","modelPresent","USAGE","CONFIG","parse","download","ctx","values","argv","name","cfg","resolveConfig","config","e","type","console","log","url","model","dir","file","into","error"],"mappings":"AAAA,SAASA,WAAW,QAAQ,qBAAqB;AACjD,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,aAAa,EAAEC,QAAQ,EAAEC,YAAY,QAAQ,uBAAuB;AAC7E,SAASC,KAAK,QAAQ,aAAa;AACnC,SAASC,MAAM,EAAEC,KAAK,QAAQ,cAAc;AAG5C,wFAAwF;AACxF,4FAA4F;AAC5F,8DAA8D;AAC9D,MAAMC,WAAoB,OAAOC;IAC/B,MAAM,EAAEC,MAAM,EAAE,GAAGH,MAAME,IAAIE,IAAI,EAAE,CAAC,OAAO,EAAEF,IAAIG,IAAI,CAAC,CAAC,EAAEP,MAAMG,QAAQ,EAAE,EAAE;QAAE,GAAGF,MAAM;IAAC;IACvF,MAAMO,MAAMJ,IAAIK,aAAa,CAACJ,OAAOK,MAAM;IAC3C,MAAMC,IAAIhB,YAAYa;IACtB,IAAI,CAACG,GAAG,MAAM,IAAIf,WAAW,kBAAkB;IAE/C,IAAIe,EAAEC,IAAI,KAAK,OAAO;QACpBC,QAAQC,GAAG,CAAC,CAAC,qBAAqB,EAAEH,EAAEI,GAAG,CAAC,oCAAoC,EAAEX,IAAIG,IAAI,CAAC,mCAAmC,CAAC;QAC7H;IACF;IACA,IAAIR,aAAaS,MAAM;QACrBK,QAAQC,GAAG,CAAC,GAAGH,EAAEK,KAAK,CAAC,yBAAyB,EAAElB,SAASa,EAAEK,KAAK,GAAG;QACrE;IACF;IACA,MAAMC,MAAM,MAAMpB,cAAcc,EAAEK,KAAK,EAAE,CAACE,MAAMC,OAASN,QAAQO,KAAK,CAAC,CAAC,SAAS,EAAET,EAAEK,KAAK,CAAC,CAAC,EAAEE,KAAK,MAAM,EAAEC,MAAM;IACjHN,QAAQC,GAAG,CAAC,CAAC,WAAW,EAAEH,EAAEK,KAAK,CAAC,MAAM,EAAEC,KAAK;AACjD;AACA,eAAed,SAAS"}
+{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/download.ts"],"sourcesContent":["import { basename } from 'node:path';\nimport { embedConfig } from '../config/index.ts';\nimport { downloadModel, isDownloadable, modelDir, modelPresent } from '../embed/store.ts';\nimport { SenseError } from '../errors.ts';\nimport { USAGE } from './index.ts';\nimport { CONFIG, parse } from './shared.ts';\nimport type { Command } from './types.ts';\n\n// A prefetch over the store's own fetch path: a search would fetch the same weights\n// lazily, so this only moves the wait earlier. Idempotent; asks nothing, so CI can run it unattended.\nconst download: Command = async (ctx) => {\n  const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.download}`, { ...CONFIG });\n  const cfg = ctx.resolveConfig(values.config as string | undefined);\n  const e = embedConfig(cfg);\n  if (!e) throw new SenseError('EMBED_DISABLED', 'no preset asks for vectors, so there is no model to download');\n\n  if (e.provider !== 'static') {\n    console.log(`embed provider is \"${e.provider}\" (${e.url}), so there is nothing to download; ${ctx.name} search reaches it over the network`);\n    return;\n  }\n  if (modelPresent(cfg)) {\n    const at = isDownloadable(e.model) ? `@${basename(modelDir(e.model))} ` : ' ';\n    console.log(`${e.model}${at}is already available in ${modelDir(e.model)}`);\n    return;\n  }\n  const dir = await downloadModel(e.model, (file, into) => console.error(`fetching ${e.model}/${file} into ${into}`));\n  console.log(`downloaded ${e.model}@${basename(dir)} into ${dir}`);\n};\nexport default download;\n"],"names":["basename","embedConfig","downloadModel","isDownloadable","modelDir","modelPresent","SenseError","USAGE","CONFIG","parse","download","ctx","values","argv","name","cfg","resolveConfig","config","e","provider","console","log","url","at","model","dir","file","into","error"],"mappings":"AAAA,SAASA,QAAQ,QAAQ,YAAY;AACrC,SAASC,WAAW,QAAQ,qBAAqB;AACjD,SAASC,aAAa,EAAEC,cAAc,EAAEC,QAAQ,EAAEC,YAAY,QAAQ,oBAAoB;AAC1F,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,KAAK,QAAQ,aAAa;AACnC,SAASC,MAAM,EAAEC,KAAK,QAAQ,cAAc;AAG5C,oFAAoF;AACpF,sGAAsG;AACtG,MAAMC,WAAoB,OAAOC;IAC/B,MAAM,EAAEC,MAAM,EAAE,GAAGH,MAAME,IAAIE,IAAI,EAAE,CAAC,OAAO,EAAEF,IAAIG,IAAI,CAAC,CAAC,EAAEP,MAAMG,QAAQ,EAAE,EAAE;QAAE,GAAGF,MAAM;IAAC;IACvF,MAAMO,MAAMJ,IAAIK,aAAa,CAACJ,OAAOK,MAAM;IAC3C,MAAMC,IAAIjB,YAAYc;IACtB,IAAI,CAACG,GAAG,MAAM,IAAIZ,WAAW,kBAAkB;IAE/C,IAAIY,EAAEC,QAAQ,KAAK,UAAU;QAC3BC,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAEH,EAAEC,QAAQ,CAAC,GAAG,EAAED,EAAEI,GAAG,CAAC,oCAAoC,EAAEX,IAAIG,IAAI,CAAC,mCAAmC,CAAC;QAC3I;IACF;IACA,IAAIT,aAAaU,MAAM;QACrB,MAAMQ,KAAKpB,eAAee,EAAEM,KAAK,IAAI,CAAC,CAAC,EAAExB,SAASI,SAASc,EAAEM,KAAK,GAAG,CAAC,CAAC,GAAG;QAC1EJ,QAAQC,GAAG,CAAC,GAAGH,EAAEM,KAAK,GAAGD,GAAG,wBAAwB,EAAEnB,SAASc,EAAEM,KAAK,GAAG;QACzE;IACF;IACA,MAAMC,MAAM,MAAMvB,cAAcgB,EAAEM,KAAK,EAAE,CAACE,MAAMC,OAASP,QAAQQ,KAAK,CAAC,CAAC,SAAS,EAAEV,EAAEM,KAAK,CAAC,CAAC,EAAEE,KAAK,MAAM,EAAEC,MAAM;IACjHP,QAAQC,GAAG,CAAC,CAAC,WAAW,EAAEH,EAAEM,KAAK,CAAC,CAAC,EAAExB,SAASyB,KAAK,MAAM,EAAEA,KAAK;AAClE;AACA,eAAef,SAAS"}

package/dist/esm/cli/index.d.ts

@@ -1,6 +1,6 @@
 import type { Command } from './types.js';
 export declare const USAGE: {
-    readonly init: "init";
+    readonly init: "init [--model id-or-path] [--provider static|openai|cohere] [--url url]";
     readonly watch: "watch [--force] [--config path]";
     readonly status: "status [--format table|json] [--config path]";
     readonly download: "download [--config path]";

package/dist/esm/cli/index.js

@@ -2,7 +2,7 @@
 // top-level usage() and the command's own parse() error/--help output, so there is one
 // copy of each line, not two that can drift.
 export const USAGE = {
-    init: 'init',
+    init: 'init [--model id-or-path] [--provider static|openai|cohere] [--url url]',
     watch: 'watch [--force] [--config path]',
     status: 'status [--format table|json] [--config path]',
     download: 'download [--config path]',

package/dist/esm/cli/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/index.ts"],"sourcesContent":["import type { Command } from './types.ts';\n\n// Each command's usage line, minus the leading \"usage: <name> \" -- shared by cli.ts's\n// top-level usage() and the command's own parse() error/--help output, so there is one\n// copy of each line, not two that can drift.\nexport const USAGE = {\n  init: 'init',\n  watch: 'watch [--force] [--config path]',\n  status: 'status [--format table|json] [--config path]',\n  download: 'download [--config path]',\n  sql: 'sql \"<statement>\" [params...] [--preset name] [--format table|json|csv] [--config path]',\n  search: 'search \"<terms>\" [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--k n] [--format table|json|csv] [--config path]',\n  map: 'map [--format table|json] [--config path]',\n  peek: 'peek <path> [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--format table|json] [--config path]',\n  path: 'path <a> <b> [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--max-depth n] [--format table|json] [--config path]',\n  related: 'related <note> [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--k n] [--format table|json|csv] [--config path]',\n} as const;\n\n// Lazy registry: a command's imports load only when it runs, so a heavy dependency in one\n// command never taxes the others (or --version/--help). These names are reserved: a saved\n// entry in config can never shadow them.\nexport const COMMANDS: Record<string, () => Promise<{ default: Command }>> = {\n  init: () => import('./init.ts'),\n  watch: () => import('./watch.ts'),\n  status: () => import('./status.ts'),\n  download: () => import('./download.ts'),\n  sql: () => import('./sql.ts'),\n  search: () => import('./search.ts'),\n  map: () => import('./map.ts'),\n  peek: () => import('./peek.ts'),\n  path: () => import('./path.ts'),\n  related: () => import('./related.ts'),\n};\n\nexport type { Command, Ctx } from './types.ts';\n"],"names":["USAGE","init","watch","status","download","sql","search","map","peek","path","related","COMMANDS"],"mappings":"AAEA,sFAAsF;AACtF,uFAAuF;AACvF,6CAA6C;AAC7C,OAAO,MAAMA,QAAQ;IACnBC,MAAM;IACNC,OAAO;IACPC,QAAQ;IACRC,UAAU;IACVC,KAAK;IACLC,QAAQ;IACRC,KAAK;IACLC,MAAM;IACNC,MAAM;IACNC,SAAS;AACX,EAAW;AAEX,0FAA0F;AAC1F,0FAA0F;AAC1F,yCAAyC;AACzC,OAAO,MAAMC,WAAgE;IAC3EV,MAAM,IAAM,MAAM,CAAC;IACnBC,OAAO,IAAM,MAAM,CAAC;IACpBC,QAAQ,IAAM,MAAM,CAAC;IACrBC,UAAU,IAAM,MAAM,CAAC;IACvBC,KAAK,IAAM,MAAM,CAAC;IAClBC,QAAQ,IAAM,MAAM,CAAC;IACrBC,KAAK,IAAM,MAAM,CAAC;IAClBC,MAAM,IAAM,MAAM,CAAC;IACnBC,MAAM,IAAM,MAAM,CAAC;IACnBC,SAAS,IAAM,MAAM,CAAC;AACxB,EAAE"}
+{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/index.ts"],"sourcesContent":["import type { Command } from './types.ts';\n\n// Each command's usage line, minus the leading \"usage: <name> \" -- shared by cli.ts's\n// top-level usage() and the command's own parse() error/--help output, so there is one\n// copy of each line, not two that can drift.\nexport const USAGE = {\n  init: 'init [--model id-or-path] [--provider static|openai|cohere] [--url url]',\n  watch: 'watch [--force] [--config path]',\n  status: 'status [--format table|json] [--config path]',\n  download: 'download [--config path]',\n  sql: 'sql \"<statement>\" [params...] [--preset name] [--format table|json|csv] [--config path]',\n  search: 'search \"<terms>\" [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--k n] [--format table|json|csv] [--config path]',\n  map: 'map [--format table|json] [--config path]',\n  peek: 'peek <path> [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--format table|json] [--config path]',\n  path: 'path <a> <b> [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--max-depth n] [--format table|json] [--config path]',\n  related: 'related <note> [--preset name] [--include glob ...] [--exclude glob ...] [--no-exclude] [--where \"<sql>\"] [--k n] [--format table|json|csv] [--config path]',\n} as const;\n\n// Lazy registry: a command's imports load only when it runs, so a heavy dependency in one\n// command never taxes the others (or --version/--help). These names are reserved: a saved\n// entry in config can never shadow them.\nexport const COMMANDS: Record<string, () => Promise<{ default: Command }>> = {\n  init: () => import('./init.ts'),\n  watch: () => import('./watch.ts'),\n  status: () => import('./status.ts'),\n  download: () => import('./download.ts'),\n  sql: () => import('./sql.ts'),\n  search: () => import('./search.ts'),\n  map: () => import('./map.ts'),\n  peek: () => import('./peek.ts'),\n  path: () => import('./path.ts'),\n  related: () => import('./related.ts'),\n};\n\nexport type { Command, Ctx } from './types.ts';\n"],"names":["USAGE","init","watch","status","download","sql","search","map","peek","path","related","COMMANDS"],"mappings":"AAEA,sFAAsF;AACtF,uFAAuF;AACvF,6CAA6C;AAC7C,OAAO,MAAMA,QAAQ;IACnBC,MAAM;IACNC,OAAO;IACPC,QAAQ;IACRC,UAAU;IACVC,KAAK;IACLC,QAAQ;IACRC,KAAK;IACLC,MAAM;IACNC,MAAM;IACNC,SAAS;AACX,EAAW;AAEX,0FAA0F;AAC1F,0FAA0F;AAC1F,yCAAyC;AACzC,OAAO,MAAMC,WAAgE;IAC3EV,MAAM,IAAM,MAAM,CAAC;IACnBC,OAAO,IAAM,MAAM,CAAC;IACpBC,QAAQ,IAAM,MAAM,CAAC;IACrBC,UAAU,IAAM,MAAM,CAAC;IACvBC,KAAK,IAAM,MAAM,CAAC;IAClBC,QAAQ,IAAM,MAAM,CAAC;IACrBC,KAAK,IAAM,MAAM,CAAC;IAClBC,MAAM,IAAM,MAAM,CAAC;IACnBC,MAAM,IAAM,MAAM,CAAC;IACnBC,SAAS,IAAM,MAAM,CAAC;AACxB,EAAE"}

package/dist/esm/cli/init.js

@@ -1,14 +1,40 @@
-import { initConfig } from '../config/index.js';
+import { DEFAULT_EMBED_MODEL, initConfig } from '../config/index.js';
 import { USAGE } from './index.js';
 import { parse } from './shared.js';
 const init = (ctx)=>{
-    parse(ctx.argv, `usage: ${ctx.name} ${USAGE.init}`, {});
-    const configPath = initConfig(process.cwd());
+    const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.init}`, {
+        model: {
+            type: 'string'
+        },
+        provider: {
+            type: 'string'
+        },
+        url: {
+            type: 'string'
+        }
+    });
+    const modelFlag = values.model;
+    const providerFlag = values.provider;
+    const model = modelFlag !== null && modelFlag !== void 0 ? modelFlag : DEFAULT_EMBED_MODEL;
+    const provider = providerFlag !== null && providerFlag !== void 0 ? providerFlag : 'static';
+    const url = values.url;
+    // An invalid --provider throws here (validateConfig's own shape error), before anything is
+    // written -- one validator, not a second copy of its rules.
+    const configPath = initConfig(process.cwd(), {
+        model: modelFlag,
+        provider: providerFlag,
+        url
+    });
     console.log(`created ${configPath}`);
     console.log(`query away: ${ctx.name} sql "SELECT path FROM frontmatter LIMIT 10"`);
-    // Semantic search needs a 124 MB model that nothing downloads implicitly; name the step
-    // here so the first search meets it as a choice rather than as a degraded result.
-    console.log(`for meaning-based search: ${ctx.name} download (one 124 MB model, once)`);
+    // Naming a model is consent to fetch it, so writing one into the config says so loudly,
+    // with the download consequence and the prefetch command, right where it was just written.
+    if (provider === 'static') {
+        console.log(`embed model: ${model} (static); nothing downloads until the first vector search -- ${ctx.name} download fetches it now instead`);
+    } else {
+        console.log(`embed model: ${model} (${provider}${url ? ` at ${url}` : ''}); reached over the network at search time, nothing to download`);
+    }
+    console.log('a non-English tree should pick a model for its languages: see the sense-setup skill or INTEGRATIONS.md');
     // The decisions a new tree faces -- features, frontmatter conventions, note size -- are not
     // in this output; name both places that hold them, since an agent may have neither.
     console.log('features and tree design: the sense-setup skill, or schema.json');

package/dist/esm/cli/init.js.map

@@ -1 +1 @@
-{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/init.ts"],"sourcesContent":["import { initConfig } from '../config/index.ts';\nimport { USAGE } from './index.ts';\nimport { parse } from './shared.ts';\nimport type { Command } from './types.ts';\n\nconst init: Command = (ctx) => {\n  parse(ctx.argv, `usage: ${ctx.name} ${USAGE.init}`, {});\n  const configPath = initConfig(process.cwd());\n  console.log(`created ${configPath}`);\n  console.log(`query away: ${ctx.name} sql \"SELECT path FROM frontmatter LIMIT 10\"`);\n  // Semantic search needs a 124 MB model that nothing downloads implicitly; name the step\n  // here so the first search meets it as a choice rather than as a degraded result.\n  console.log(`for meaning-based search: ${ctx.name} download (one 124 MB model, once)`);\n  // The decisions a new tree faces -- features, frontmatter conventions, note size -- are not\n  // in this output; name both places that hold them, since an agent may have neither.\n  console.log('features and tree design: the sense-setup skill, or schema.json');\n};\nexport default init;\n"],"names":["initConfig","USAGE","parse","init","ctx","argv","name","configPath","process","cwd","console","log"],"mappings":"AAAA,SAASA,UAAU,QAAQ,qBAAqB;AAChD,SAASC,KAAK,QAAQ,aAAa;AACnC,SAASC,KAAK,QAAQ,cAAc;AAGpC,MAAMC,OAAgB,CAACC;IACrBF,MAAME,IAAIC,IAAI,EAAE,CAAC,OAAO,EAAED,IAAIE,IAAI,CAAC,CAAC,EAAEL,MAAME,IAAI,EAAE,EAAE,CAAC;IACrD,MAAMI,aAAaP,WAAWQ,QAAQC,GAAG;IACzCC,QAAQC,GAAG,CAAC,CAAC,QAAQ,EAAEJ,YAAY;IACnCG,QAAQC,GAAG,CAAC,CAAC,YAAY,EAAEP,IAAIE,IAAI,CAAC,4CAA4C,CAAC;IACjF,wFAAwF;IACxF,kFAAkF;IAClFI,QAAQC,GAAG,CAAC,CAAC,0BAA0B,EAAEP,IAAIE,IAAI,CAAC,kCAAkC,CAAC;IACrF,4FAA4F;IAC5F,oFAAoF;IACpFI,QAAQC,GAAG,CAAC;AACd;AACA,eAAeR,KAAK"}
+{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/init.ts"],"sourcesContent":["import { DEFAULT_EMBED_MODEL, initConfig } from '../config/index.ts';\nimport { USAGE } from './index.ts';\nimport { parse } from './shared.ts';\nimport type { Command } from './types.ts';\n\nconst init: Command = (ctx) => {\n  const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.init}`, {\n    model: { type: 'string' },\n    provider: { type: 'string' },\n    url: { type: 'string' },\n  });\n  const modelFlag = values.model as string | undefined;\n  const providerFlag = values.provider as string | undefined;\n  const model = modelFlag ?? DEFAULT_EMBED_MODEL;\n  const provider = providerFlag ?? 'static';\n  const url = values.url as string | undefined;\n  // An invalid --provider throws here (validateConfig's own shape error), before anything is\n  // written -- one validator, not a second copy of its rules.\n  const configPath = initConfig(process.cwd(), { model: modelFlag, provider: providerFlag, url });\n  console.log(`created ${configPath}`);\n  console.log(`query away: ${ctx.name} sql \"SELECT path FROM frontmatter LIMIT 10\"`);\n  // Naming a model is consent to fetch it, so writing one into the config says so loudly,\n  // with the download consequence and the prefetch command, right where it was just written.\n  if (provider === 'static') {\n    console.log(`embed model: ${model} (static); nothing downloads until the first vector search -- ${ctx.name} download fetches it now instead`);\n  } else {\n    console.log(`embed model: ${model} (${provider}${url ? ` at ${url}` : ''}); reached over the network at search time, nothing to download`);\n  }\n  console.log('a non-English tree should pick a model for its languages: see the sense-setup skill or INTEGRATIONS.md');\n  // The decisions a new tree faces -- features, frontmatter conventions, note size -- are not\n  // in this output; name both places that hold them, since an agent may have neither.\n  console.log('features and tree design: the sense-setup skill, or schema.json');\n};\nexport default init;\n"],"names":["DEFAULT_EMBED_MODEL","initConfig","USAGE","parse","init","ctx","values","argv","name","model","type","provider","url","modelFlag","providerFlag","configPath","process","cwd","console","log"],"mappings":"AAAA,SAASA,mBAAmB,EAAEC,UAAU,QAAQ,qBAAqB;AACrE,SAASC,KAAK,QAAQ,aAAa;AACnC,SAASC,KAAK,QAAQ,cAAc;AAGpC,MAAMC,OAAgB,CAACC;IACrB,MAAM,EAAEC,MAAM,EAAE,GAAGH,MAAME,IAAIE,IAAI,EAAE,CAAC,OAAO,EAAEF,IAAIG,IAAI,CAAC,CAAC,EAAEN,MAAME,IAAI,EAAE,EAAE;QACrEK,OAAO;YAAEC,MAAM;QAAS;QACxBC,UAAU;YAAED,MAAM;QAAS;QAC3BE,KAAK;YAAEF,MAAM;QAAS;IACxB;IACA,MAAMG,YAAYP,OAAOG,KAAK;IAC9B,MAAMK,eAAeR,OAAOK,QAAQ;IACpC,MAAMF,QAAQI,sBAAAA,uBAAAA,YAAab;IAC3B,MAAMW,WAAWG,yBAAAA,0BAAAA,eAAgB;IACjC,MAAMF,MAAMN,OAAOM,GAAG;IACtB,2FAA2F;IAC3F,4DAA4D;IAC5D,MAAMG,aAAad,WAAWe,QAAQC,GAAG,IAAI;QAAER,OAAOI;QAAWF,UAAUG;QAAcF;IAAI;IAC7FM,QAAQC,GAAG,CAAC,CAAC,QAAQ,EAAEJ,YAAY;IACnCG,QAAQC,GAAG,CAAC,CAAC,YAAY,EAAEd,IAAIG,IAAI,CAAC,4CAA4C,CAAC;IACjF,wFAAwF;IACxF,2FAA2F;IAC3F,IAAIG,aAAa,UAAU;QACzBO,QAAQC,GAAG,CAAC,CAAC,aAAa,EAAEV,MAAM,8DAA8D,EAAEJ,IAAIG,IAAI,CAAC,gCAAgC,CAAC;IAC9I,OAAO;QACLU,QAAQC,GAAG,CAAC,CAAC,aAAa,EAAEV,MAAM,EAAE,EAAEE,WAAWC,MAAM,CAAC,IAAI,EAAEA,KAAK,GAAG,GAAG,+DAA+D,CAAC;IAC3I;IACAM,QAAQC,GAAG,CAAC;IACZ,4FAA4F;IAC5F,oFAAoF;IACpFD,QAAQC,GAAG,CAAC;AACd;AACA,eAAef,KAAK"}

package/dist/esm/cli/status.js

@@ -2,7 +2,9 @@ import { dirname } from 'node:path';
 import { presetCoverage } from '../commands/index.js';
 import { embedConfig, featureStates, SUPPORTED_CONFIG_VERSION } from '../config/index.js';
 import { docCount, getMeta, open, SCHEMA_VERSION } from '../db/index.js';
-import { isDownloadable, MODEL_FILENAMES, modelDir, modelPresent } from '../features/embed.js';
+import { probeReachable } from '../embed/http.js';
+import { languageDistribution } from '../embed/langfit.js';
+import { isDownloadable, MODEL_FILENAMES, modelDir, modelPresent, readLanguages } from '../embed/store.js';
 import { featuresLine, presetsLines } from '../output.js';
 import { USAGE } from './index.js';
 import { CONFIG, FORMAT, formatOf, parse, printWarnings } from './shared.js';
@@ -38,8 +40,16 @@ export function redactUrl(url) {
         return url;
     }
 }
-const status = (ctx)=>{
-    var _process_env_e_key, _cfg_migratedFrom, _cfg_unknownKeys, _cfg_queries;
+// A stopped Ollama/LM Studio/Cohere endpoint prints identically to a running one otherwise; any
+// HTTP response counts as reachable (probeReachable), so this only distinguishes up from down.
+async function embedReachable(e) {
+    var _e_url;
+    if (e.provider === 'static' || e.provider === 'openai' && !e.url) return null;
+    const base = ((_e_url = e.url) !== null && _e_url !== void 0 ? _e_url : 'https://api.cohere.com').replace(/\/+$/, '');
+    return probeReachable(e.provider === 'openai' ? `${base}/models` : base);
+}
+const status = async (ctx)=>{
+    var _languageDistribution, _process_env_e_key, _cfg_migratedFrom, _cfg_unknownKeys, _cfg_queries;
     const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.status}`, {
         ...FORMAT,
         ...CONFIG
@@ -51,8 +61,12 @@ const status = (ctx)=>{
     const features = featureStates(cfg);
     const e = embedConfig(cfg);
     const hasModel = e ? modelPresent(cfg) : false;
+    const reachable = e ? await embedReachable(e) : null;
     const vectors = e ? vectorState(db) : null;
     const presets = presetCoverage(db, cfg);
+    const staticLanguages = e && e.provider === 'static' ? readLanguages(e.model) : undefined;
+    const declaredLanguages = staticLanguages && staticLanguages.length > 0 ? staticLanguages : null;
+    const detectedLanguages = (_languageDistribution = languageDistribution(db)) !== null && _languageDistribution !== void 0 ? _languageDistribution : null;
     const heartbeat = getMeta(db, 'watch_heartbeat');
     // Derived at open() (F): 3x the largest reconcile this cache has ever held its write
     // transaction for, floored at 30s -- read back from the connection rather than
@@ -77,12 +91,15 @@ const status = (ctx)=>{
         features: features.on,
         featuresOff: features.off,
         embed: e ? {
-            type: e.type,
+            provider: e.provider,
             model: e.model,
-            dir: e.type === 'static' ? modelDir(e.model) : null,
+            dir: e.provider === 'static' ? modelDir(e.model) : null,
             url: e.url ? redactUrl(e.url) : null,
+            reachable,
             keyEnv,
             downloaded: hasModel,
+            languages: declaredLanguages,
+            detectedLanguages,
             ...vectors !== null && vectors !== void 0 ? vectors : {}
         } : null,
         presets,
@@ -107,12 +124,23 @@ const status = (ctx)=>{
         // not as an empty corpus.
         if (!e) console.log('embed:    off (no preset asks for vectors)');
         else {
-            console.log(`embed:    ${e.type} ${e.model}`);
-            if (e.type === 'static') console.log(`          model:   ${modelDir(e.model)} (${hasModel ? 'present' : `missing, ${isDownloadable(e.model) ? `run \`${ctx.name} download\`` : `no ${MODEL_FILENAMES} in that directory`}`})`);
-            if (e.url) console.log(`          url:     ${redactUrl(e.url)}`);
+            console.log(`embed:    ${e.provider} ${e.model}`);
+            if (e.provider === 'static') console.log(`          model:   ${modelDir(e.model)} (${hasModel ? 'present' : `missing, ${isDownloadable(e.model) ? `fetches automatically at first search, or run \`${ctx.name} download\` to prefetch` : `no ${MODEL_FILENAMES} in that directory`}`})`);
+            if (declaredLanguages) console.log(`          languages: ${declaredLanguages.join(', ')} (declared by the model card)`);
+            // Cohere defaults to api.cohere.com with no url in the config; show it anyway so the
+            // reachability probe below has something to name.
+            if (e.url || reachable !== null) {
+                const shown = e.url ? redactUrl(e.url) : 'https://api.cohere.com (default)';
+                console.log(`          url:     ${shown}${reachable === null ? '' : reachable ? ' (reachable)' : ' (unreachable)'}`);
+            }
             if (keyEnv) console.log(`          key:     env ${keyEnv.name} (${keyEnv.set ? 'set' : 'NOT SET'})`);
             if (!hasModel) console.log('          vectors: off until the model is present; search runs on words and links');
             else if (vectors) console.log(`          vectors: ${vectors.embedded} embedded, ${vectors.pending} pending (embedded on the first semantic search)`);
+            if (detectedLanguages) {
+                const total = Object.values(detectedLanguages).reduce((a, b)=>a + b, 0);
+                const breakdown = Object.entries(detectedLanguages).sort((a, b)=>b[1] - a[1]).map(([code, n])=>`${code} ${Math.round(n / total * 100)}%`).join(', ');
+                console.log(`          detected: ${breakdown} (${total} chunks classified)`);
+            }
         }
         console.log('');
         for (const line of presetsLines(presets))console.log(line);

package/dist/esm/cli/status.js.map

@@ -1 +1 @@
-{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/status.ts"],"sourcesContent":["import { dirname } from 'node:path';\nimport { presetCoverage } from '../commands/index.ts';\nimport { embedConfig, featureStates, SUPPORTED_CONFIG_VERSION } from '../config/index.ts';\nimport { docCount, getMeta, open, SCHEMA_VERSION } from '../db/index.ts';\nimport { isDownloadable, MODEL_FILENAMES, modelDir, modelPresent } from '../features/embed.ts';\nimport { featuresLine, presetsLines } from '../output.ts';\nimport { USAGE } from './index.ts';\nimport { CONFIG, FORMAT, formatOf, parse, printWarnings } from './shared.ts';\nimport type { Command } from './types.ts';\n\n// status names what you cannot read off the config file: locations sense chose, values it\n// derived, state it stored. Config content is not echoed -- the config line says which file to\n// read, and duplicating its globs here would create two places that disagree after an edit.\n// The exception is the embed block's url and key, whose effect is remote and otherwise invisible.\n\n// Vectors are computed on the first semantic query, not at reconcile, so a tree can have a\n// semantic-on preset with nothing embedded yet -- reported here because a fast reconcile on\n// such a tree otherwise reads as \"the feature isn't on\".\nfunction vectorState(db: ReturnType<typeof open>['db']): { embedded: number; pending: number } | null {\n  try {\n    const row = db.prepare('SELECT COUNT(vector) AS embedded, COUNT(*) - COUNT(vector) AS pending FROM embeddings').get() as { embedded: number; pending: number };\n    return row;\n  } catch {\n    return null;\n  }\n}\n\nfunction parseErrorCount(db: ReturnType<typeof open>['db']): number {\n  const row = db.prepare(`SELECT COUNT(*) AS n FROM frontmatter WHERE \"_parse_error\" IS NOT NULL`).get() as { n: number };\n  return row.n;\n}\n\n// A URL can carry credentials in its userinfo (https://user:pass@host/v1). status output gets\n// pasted into issues and transcripts, so it is always safe to paste.\nexport function redactUrl(url: string): string {\n  try {\n    const parsed = new URL(url);\n    if (!parsed.username && !parsed.password) return url;\n    parsed.username = '';\n    parsed.password = '';\n    return `${parsed.toString().replace(/\\/$/, '')} (credentials redacted)`;\n  } catch {\n    return url;\n  }\n}\n\nconst status: Command = (ctx) => {\n  const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.status}`, { ...FORMAT, ...CONFIG });\n  const format = formatOf(values);\n  const cfg = ctx.resolveConfig(values.config as string | undefined);\n  const { db, dbPath, warnings } = open(cfg);\n  printWarnings(warnings);\n\n  const features = featureStates(cfg);\n  const e = embedConfig(cfg);\n  const hasModel = e ? modelPresent(cfg) : false;\n  const vectors = e ? vectorState(db) : null;\n  const presets = presetCoverage(db, cfg);\n  const heartbeat = getMeta(db, 'watch_heartbeat');\n  // Derived at open() (F): 3x the largest reconcile this cache has ever held its write\n  // transaction for, floored at 30s -- read back from the connection rather than\n  // recomputed here, so this always reports what open() actually set.\n  const busyTimeoutMs = (db.prepare('PRAGMA busy_timeout').get() as { timeout: number }).timeout;\n  // The env var holds the token; only its name and whether it is set are ever reported.\n  const keyEnv = e?.key ? { name: e.key, set: (process.env[e.key] ?? '') !== '' } : null;\n  const result = {\n    config: cfg.configPath,\n    configVersion: SUPPORTED_CONFIG_VERSION,\n    migratedFrom: cfg.migratedFrom ?? null,\n    unknownConfigKeys: cfg.unknownKeys ?? null,\n    tree: cfg.baseDir,\n    cache: dirname(dbPath),\n    db: dbPath,\n    cacheSchema: SCHEMA_VERSION,\n    docs: docCount(db),\n    unparseableFrontmatter: parseErrorCount(db),\n    features: features.on,\n    featuresOff: features.off,\n    embed: e ? { type: e.type, model: e.model, dir: e.type === 'static' ? modelDir(e.model) : null, url: e.url ? redactUrl(e.url) : null, keyEnv, downloaded: hasModel, ...(vectors ?? {}) } : null,\n    presets,\n    queries: Object.keys(cfg.queries ?? {}).length,\n    watcherPid: getMeta(db, 'watch_pid'),\n    watcherHeartbeatSecondsAgo: heartbeat ? Math.round((Date.now() - Date.parse(heartbeat)) / 1000) : null,\n    busyTimeoutMs,\n  };\n\n  if (format === 'json') {\n    console.log(JSON.stringify(result, null, 2));\n  } else {\n    const migrated = cfg.migratedFrom !== undefined ? `, migrated from v${cfg.migratedFrom} on this run` : '';\n    console.log(`config:   ${result.config} (v${SUPPORTED_CONFIG_VERSION}${migrated})`);\n    // A load-time stderr warning has scrolled away by the time anyone looks here.\n    if (cfg.unknownKeys) console.log(`          unknown keys, ignored: ${cfg.unknownKeys.join(', ')}`);\n    console.log(`tree:     ${result.tree}`);\n    console.log(`cache:    ${result.cache} (schema ${SCHEMA_VERSION}; delete this directory to reset)`);\n    console.log(`docs:     ${result.docs}${result.unparseableFrontmatter > 0 ? `  (${result.unparseableFrontmatter} with unparseable frontmatter: WHERE _parse_error IS NOT NULL)` : ''}`);\n    console.log('');\n    console.log(featuresLine(features));\n    // One authoritative block, off states included: an absent model reads as a degraded search,\n    // not as an empty corpus.\n    if (!e) console.log('embed:    off (no preset asks for vectors)');\n    else {\n      console.log(`embed:    ${e.type} ${e.model}`);\n      if (e.type === 'static') console.log(`          model:   ${modelDir(e.model)} (${hasModel ? 'present' : `missing, ${isDownloadable(e.model) ? `run \\`${ctx.name} download\\`` : `no ${MODEL_FILENAMES} in that directory`}`})`);\n      if (e.url) console.log(`          url:     ${redactUrl(e.url)}`);\n      if (keyEnv) console.log(`          key:     env ${keyEnv.name} (${keyEnv.set ? 'set' : 'NOT SET'})`);\n      if (!hasModel) console.log('          vectors: off until the model is present; search runs on words and links');\n      else if (vectors) console.log(`          vectors: ${vectors.embedded} embedded, ${vectors.pending} pending (embedded on the first semantic search)`);\n    }\n    console.log('');\n    for (const line of presetsLines(presets)) console.log(line);\n    console.log('');\n    console.log(`queries:  ${result.queries} saved (${ctx.name} --list)`);\n    const pid = result.watcherPid ? `pid ${result.watcherPid}, ` : '';\n    console.log(result.watcherHeartbeatSecondsAgo === null ? 'watcher:  none' : `watcher:  ${pid}last heartbeat ${result.watcherHeartbeatSecondsAgo}s ago`);\n    console.log(`sqlite:   busy_timeout ${result.busyTimeoutMs}ms (derived: 3x the largest reconcile this cache has recorded, floored at 30000ms)`);\n  }\n  db.close();\n};\nexport default status;\n"],"names":["dirname","presetCoverage","embedConfig","featureStates","SUPPORTED_CONFIG_VERSION","docCount","getMeta","open","SCHEMA_VERSION","isDownloadable","MODEL_FILENAMES","modelDir","modelPresent","featuresLine","presetsLines","USAGE","CONFIG","FORMAT","formatOf","parse","printWarnings","vectorState","db","row","prepare","get","parseErrorCount","n","redactUrl","url","parsed","URL","username","password","toString","replace","status","ctx","process","cfg","values","argv","name","format","resolveConfig","config","dbPath","warnings","features","e","hasModel","vectors","presets","heartbeat","busyTimeoutMs","timeout","keyEnv","key","set","env","result","configPath","configVersion","migratedFrom","unknownConfigKeys","unknownKeys","tree","baseDir","cache","cacheSchema","docs","unparseableFrontmatter","on","featuresOff","off","embed","type","model","dir","downloaded","queries","Object","keys","length","watcherPid","watcherHeartbeatSecondsAgo","Math","round","Date","now","console","log","JSON","stringify","migrated","undefined","join","embedded","pending","line","pid","close"],"mappings":"AAAA,SAASA,OAAO,QAAQ,YAAY;AACpC,SAASC,cAAc,QAAQ,uBAAuB;AACtD,SAASC,WAAW,EAAEC,aAAa,EAAEC,wBAAwB,QAAQ,qBAAqB;AAC1F,SAASC,QAAQ,EAAEC,OAAO,EAAEC,IAAI,EAAEC,cAAc,QAAQ,iBAAiB;AACzE,SAASC,cAAc,EAAEC,eAAe,EAAEC,QAAQ,EAAEC,YAAY,QAAQ,uBAAuB;AAC/F,SAASC,YAAY,EAAEC,YAAY,QAAQ,eAAe;AAC1D,SAASC,KAAK,QAAQ,aAAa;AACnC,SAASC,MAAM,EAAEC,MAAM,EAAEC,QAAQ,EAAEC,KAAK,EAAEC,aAAa,QAAQ,cAAc;AAG7E,0FAA0F;AAC1F,+FAA+F;AAC/F,4FAA4F;AAC5F,kGAAkG;AAElG,2FAA2F;AAC3F,4FAA4F;AAC5F,yDAAyD;AACzD,SAASC,YAAYC,EAAiC;IACpD,IAAI;QACF,MAAMC,MAAMD,GAAGE,OAAO,CAAC,yFAAyFC,GAAG;QACnH,OAAOF;IACT,EAAE,OAAM;QACN,OAAO;IACT;AACF;AAEA,SAASG,gBAAgBJ,EAAiC;IACxD,MAAMC,MAAMD,GAAGE,OAAO,CAAC,CAAC,sEAAsE,CAAC,EAAEC,GAAG;IACpG,OAAOF,IAAII,CAAC;AACd;AAEA,8FAA8F;AAC9F,qEAAqE;AACrE,OAAO,SAASC,UAAUC,GAAW;IACnC,IAAI;QACF,MAAMC,SAAS,IAAIC,IAAIF;QACvB,IAAI,CAACC,OAAOE,QAAQ,IAAI,CAACF,OAAOG,QAAQ,EAAE,OAAOJ;QACjDC,OAAOE,QAAQ,GAAG;QAClBF,OAAOG,QAAQ,GAAG;QAClB,OAAO,GAAGH,OAAOI,QAAQ,GAAGC,OAAO,CAAC,OAAO,IAAI,uBAAuB,CAAC;IACzE,EAAE,OAAM;QACN,OAAON;IACT;AACF;AAEA,MAAMO,SAAkB,CAACC;QAkBsBC,oBAI7BC,mBACKA,kBAWEA;IAjCvB,MAAM,EAAEC,MAAM,EAAE,GAAGrB,MAAMkB,IAAII,IAAI,EAAE,CAAC,OAAO,EAAEJ,IAAIK,IAAI,CAAC,CAAC,EAAE3B,MAAMqB,MAAM,EAAE,EAAE;QAAE,GAAGnB,MAAM;QAAE,GAAGD,MAAM;IAAC;IAChG,MAAM2B,SAASzB,SAASsB;IACxB,MAAMD,MAAMF,IAAIO,aAAa,CAACJ,OAAOK,MAAM;IAC3C,MAAM,EAAEvB,EAAE,EAAEwB,MAAM,EAAEC,QAAQ,EAAE,GAAGxC,KAAKgC;IACtCnB,cAAc2B;IAEd,MAAMC,WAAW7C,cAAcoC;IAC/B,MAAMU,IAAI/C,YAAYqC;IACtB,MAAMW,WAAWD,IAAIrC,aAAa2B,OAAO;IACzC,MAAMY,UAAUF,IAAI5B,YAAYC,MAAM;IACtC,MAAM8B,UAAUnD,eAAeqB,IAAIiB;IACnC,MAAMc,YAAY/C,QAAQgB,IAAI;IAC9B,qFAAqF;IACrF,+EAA+E;IAC/E,oEAAoE;IACpE,MAAMgC,gBAAgB,AAAChC,GAAGE,OAAO,CAAC,uBAAuBC,GAAG,GAA2B8B,OAAO;IAC9F,sFAAsF;IACtF,MAAMC,SAASP,CAAAA,cAAAA,wBAAAA,EAAGQ,GAAG,IAAG;QAAEf,MAAMO,EAAEQ,GAAG;QAAEC,KAAK,EAACpB,qBAAAA,QAAQqB,GAAG,CAACV,EAAEQ,GAAG,CAAC,cAAlBnB,gCAAAA,qBAAsB,QAAQ;IAAG,IAAI;IAClF,MAAMsB,SAAS;QACbf,QAAQN,IAAIsB,UAAU;QACtBC,eAAe1D;QACf2D,YAAY,GAAExB,oBAAAA,IAAIwB,YAAY,cAAhBxB,+BAAAA,oBAAoB;QAClCyB,iBAAiB,GAAEzB,mBAAAA,IAAI0B,WAAW,cAAf1B,8BAAAA,mBAAmB;QACtC2B,MAAM3B,IAAI4B,OAAO;QACjBC,OAAOpE,QAAQ8C;QACfxB,IAAIwB;QACJuB,aAAa7D;QACb8D,MAAMjE,SAASiB;QACfiD,wBAAwB7C,gBAAgBJ;QACxC0B,UAAUA,SAASwB,EAAE;QACrBC,aAAazB,SAAS0B,GAAG;QACzBC,OAAO1B,IAAI;YAAE2B,MAAM3B,EAAE2B,IAAI;YAAEC,OAAO5B,EAAE4B,KAAK;YAAEC,KAAK7B,EAAE2B,IAAI,KAAK,WAAWjE,SAASsC,EAAE4B,KAAK,IAAI;YAAMhD,KAAKoB,EAAEpB,GAAG,GAAGD,UAAUqB,EAAEpB,GAAG,IAAI;YAAM2B;YAAQuB,YAAY7B;eAAcC,oBAAAA,qBAAAA,UAAW,CAAC,CAAhB;QAAmB,IAAI;QAC3LC;QACA4B,SAASC,OAAOC,IAAI,EAAC3C,eAAAA,IAAIyC,OAAO,cAAXzC,0BAAAA,eAAe,CAAC,GAAG4C,MAAM;QAC9CC,YAAY9E,QAAQgB,IAAI;QACxB+D,4BAA4BhC,YAAYiC,KAAKC,KAAK,CAAC,AAACC,CAAAA,KAAKC,GAAG,KAAKD,KAAKrE,KAAK,CAACkC,UAAS,IAAK,QAAQ;QAClGC;IACF;IAEA,IAAIX,WAAW,QAAQ;QACrB+C,QAAQC,GAAG,CAACC,KAAKC,SAAS,CAACjC,QAAQ,MAAM;IAC3C,OAAO;QACL,MAAMkC,WAAWvD,IAAIwB,YAAY,KAAKgC,YAAY,CAAC,iBAAiB,EAAExD,IAAIwB,YAAY,CAAC,YAAY,CAAC,GAAG;QACvG2B,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE/B,OAAOf,MAAM,CAAC,GAAG,EAAEzC,2BAA2B0F,SAAS,CAAC,CAAC;QAClF,8EAA8E;QAC9E,IAAIvD,IAAI0B,WAAW,EAAEyB,QAAQC,GAAG,CAAC,CAAC,iCAAiC,EAAEpD,IAAI0B,WAAW,CAAC+B,IAAI,CAAC,OAAO;QACjGN,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE/B,OAAOM,IAAI,EAAE;QACtCwB,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE/B,OAAOQ,KAAK,CAAC,SAAS,EAAE5D,eAAe,iCAAiC,CAAC;QAClGkF,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE/B,OAAOU,IAAI,GAAGV,OAAOW,sBAAsB,GAAG,IAAI,CAAC,GAAG,EAAEX,OAAOW,sBAAsB,CAAC,8DAA8D,CAAC,GAAG,IAAI;QACrLmB,QAAQC,GAAG,CAAC;QACZD,QAAQC,GAAG,CAAC9E,aAAamC;QACzB,4FAA4F;QAC5F,0BAA0B;QAC1B,IAAI,CAACC,GAAGyC,QAAQC,GAAG,CAAC;aACf;YACHD,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE1C,EAAE2B,IAAI,CAAC,CAAC,EAAE3B,EAAE4B,KAAK,EAAE;YAC5C,IAAI5B,EAAE2B,IAAI,KAAK,UAAUc,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAEhF,SAASsC,EAAE4B,KAAK,EAAE,EAAE,EAAE3B,WAAW,YAAY,CAAC,SAAS,EAAEzC,eAAewC,EAAE4B,KAAK,IAAI,CAAC,MAAM,EAAExC,IAAIK,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,GAAG,EAAEhC,gBAAgB,kBAAkB,CAAC,EAAE,CAAC,CAAC,CAAC;YAC7N,IAAIuC,EAAEpB,GAAG,EAAE6D,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAE/D,UAAUqB,EAAEpB,GAAG,GAAG;YAC/D,IAAI2B,QAAQkC,QAAQC,GAAG,CAAC,CAAC,uBAAuB,EAAEnC,OAAOd,IAAI,CAAC,EAAE,EAAEc,OAAOE,GAAG,GAAG,QAAQ,UAAU,CAAC,CAAC;YACnG,IAAI,CAACR,UAAUwC,QAAQC,GAAG,CAAC;iBACtB,IAAIxC,SAASuC,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAExC,QAAQ8C,QAAQ,CAAC,WAAW,EAAE9C,QAAQ+C,OAAO,CAAC,gDAAgD,CAAC;QACrJ;QACAR,QAAQC,GAAG,CAAC;QACZ,KAAK,MAAMQ,QAAQrF,aAAasC,SAAUsC,QAAQC,GAAG,CAACQ;QACtDT,QAAQC,GAAG,CAAC;QACZD,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE/B,OAAOoB,OAAO,CAAC,QAAQ,EAAE3C,IAAIK,IAAI,CAAC,QAAQ,CAAC;QACpE,MAAM0D,MAAMxC,OAAOwB,UAAU,GAAG,CAAC,IAAI,EAAExB,OAAOwB,UAAU,CAAC,EAAE,CAAC,GAAG;QAC/DM,QAAQC,GAAG,CAAC/B,OAAOyB,0BAA0B,KAAK,OAAO,mBAAmB,CAAC,UAAU,EAAEe,IAAI,eAAe,EAAExC,OAAOyB,0BAA0B,CAAC,KAAK,CAAC;QACtJK,QAAQC,GAAG,CAAC,CAAC,uBAAuB,EAAE/B,OAAON,aAAa,CAAC,kFAAkF,CAAC;IAChJ;IACAhC,GAAG+E,KAAK;AACV;AACA,eAAejE,OAAO"}
+{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/status.ts"],"sourcesContent":["import { dirname } from 'node:path';\nimport { presetCoverage } from '../commands/index.ts';\nimport { embedConfig, featureStates, SUPPORTED_CONFIG_VERSION } from '../config/index.ts';\nimport { docCount, getMeta, open, SCHEMA_VERSION } from '../db/index.ts';\nimport { probeReachable } from '../embed/http.ts';\nimport { languageDistribution } from '../embed/langfit.ts';\nimport { isDownloadable, MODEL_FILENAMES, modelDir, modelPresent, readLanguages } from '../embed/store.ts';\nimport { featuresLine, presetsLines } from '../output.ts';\nimport { USAGE } from './index.ts';\nimport { CONFIG, FORMAT, formatOf, parse, printWarnings } from './shared.ts';\nimport type { Command } from './types.ts';\n\n// status names what you cannot read off the config file: locations sense chose, values it\n// derived, state it stored. Config content is not echoed -- the config line says which file to\n// read, and duplicating its globs here would create two places that disagree after an edit.\n// The exception is the embed block's url and key, whose effect is remote and otherwise invisible.\n\n// Vectors are computed on the first semantic query, not at reconcile, so a tree can have a\n// semantic-on preset with nothing embedded yet -- reported here because a fast reconcile on\n// such a tree otherwise reads as \"the feature isn't on\".\nfunction vectorState(db: ReturnType<typeof open>['db']): { embedded: number; pending: number } | null {\n  try {\n    const row = db.prepare('SELECT COUNT(vector) AS embedded, COUNT(*) - COUNT(vector) AS pending FROM embeddings').get() as { embedded: number; pending: number };\n    return row;\n  } catch {\n    return null;\n  }\n}\n\nfunction parseErrorCount(db: ReturnType<typeof open>['db']): number {\n  const row = db.prepare(`SELECT COUNT(*) AS n FROM frontmatter WHERE \"_parse_error\" IS NOT NULL`).get() as { n: number };\n  return row.n;\n}\n\n// A URL can carry credentials in its userinfo (https://user:pass@host/v1). status output gets\n// pasted into issues and transcripts, so it is always safe to paste.\nexport function redactUrl(url: string): string {\n  try {\n    const parsed = new URL(url);\n    if (!parsed.username && !parsed.password) return url;\n    parsed.username = '';\n    parsed.password = '';\n    return `${parsed.toString().replace(/\\/$/, '')} (credentials redacted)`;\n  } catch {\n    return url;\n  }\n}\n\n// A stopped Ollama/LM Studio/Cohere endpoint prints identically to a running one otherwise; any\n// HTTP response counts as reachable (probeReachable), so this only distinguishes up from down.\nasync function embedReachable(e: { provider: string; url?: string }): Promise<boolean | null> {\n  if (e.provider === 'static' || (e.provider === 'openai' && !e.url)) return null;\n  const base = (e.url ?? 'https://api.cohere.com').replace(/\\/+$/, '');\n  return probeReachable(e.provider === 'openai' ? `${base}/models` : base);\n}\n\nconst status: Command = async (ctx) => {\n  const { values } = parse(ctx.argv, `usage: ${ctx.name} ${USAGE.status}`, { ...FORMAT, ...CONFIG });\n  const format = formatOf(values);\n  const cfg = ctx.resolveConfig(values.config as string | undefined);\n  const { db, dbPath, warnings } = open(cfg);\n  printWarnings(warnings);\n\n  const features = featureStates(cfg);\n  const e = embedConfig(cfg);\n  const hasModel = e ? modelPresent(cfg) : false;\n  const reachable = e ? await embedReachable(e) : null;\n  const vectors = e ? vectorState(db) : null;\n  const presets = presetCoverage(db, cfg);\n  const staticLanguages = e && e.provider === 'static' ? readLanguages(e.model) : undefined;\n  const declaredLanguages = staticLanguages && staticLanguages.length > 0 ? staticLanguages : null;\n  const detectedLanguages = languageDistribution(db) ?? null;\n  const heartbeat = getMeta(db, 'watch_heartbeat');\n  // Derived at open() (F): 3x the largest reconcile this cache has ever held its write\n  // transaction for, floored at 30s -- read back from the connection rather than\n  // recomputed here, so this always reports what open() actually set.\n  const busyTimeoutMs = (db.prepare('PRAGMA busy_timeout').get() as { timeout: number }).timeout;\n  // The env var holds the token; only its name and whether it is set are ever reported.\n  const keyEnv = e?.key ? { name: e.key, set: (process.env[e.key] ?? '') !== '' } : null;\n  const result = {\n    config: cfg.configPath,\n    configVersion: SUPPORTED_CONFIG_VERSION,\n    migratedFrom: cfg.migratedFrom ?? null,\n    unknownConfigKeys: cfg.unknownKeys ?? null,\n    tree: cfg.baseDir,\n    cache: dirname(dbPath),\n    db: dbPath,\n    cacheSchema: SCHEMA_VERSION,\n    docs: docCount(db),\n    unparseableFrontmatter: parseErrorCount(db),\n    features: features.on,\n    featuresOff: features.off,\n    embed: e ? { provider: e.provider, model: e.model, dir: e.provider === 'static' ? modelDir(e.model) : null, url: e.url ? redactUrl(e.url) : null, reachable, keyEnv, downloaded: hasModel, languages: declaredLanguages, detectedLanguages, ...(vectors ?? {}) } : null,\n    presets,\n    queries: Object.keys(cfg.queries ?? {}).length,\n    watcherPid: getMeta(db, 'watch_pid'),\n    watcherHeartbeatSecondsAgo: heartbeat ? Math.round((Date.now() - Date.parse(heartbeat)) / 1000) : null,\n    busyTimeoutMs,\n  };\n\n  if (format === 'json') {\n    console.log(JSON.stringify(result, null, 2));\n  } else {\n    const migrated = cfg.migratedFrom !== undefined ? `, migrated from v${cfg.migratedFrom} on this run` : '';\n    console.log(`config:   ${result.config} (v${SUPPORTED_CONFIG_VERSION}${migrated})`);\n    // A load-time stderr warning has scrolled away by the time anyone looks here.\n    if (cfg.unknownKeys) console.log(`          unknown keys, ignored: ${cfg.unknownKeys.join(', ')}`);\n    console.log(`tree:     ${result.tree}`);\n    console.log(`cache:    ${result.cache} (schema ${SCHEMA_VERSION}; delete this directory to reset)`);\n    console.log(`docs:     ${result.docs}${result.unparseableFrontmatter > 0 ? `  (${result.unparseableFrontmatter} with unparseable frontmatter: WHERE _parse_error IS NOT NULL)` : ''}`);\n    console.log('');\n    console.log(featuresLine(features));\n    // One authoritative block, off states included: an absent model reads as a degraded search,\n    // not as an empty corpus.\n    if (!e) console.log('embed:    off (no preset asks for vectors)');\n    else {\n      console.log(`embed:    ${e.provider} ${e.model}`);\n      if (e.provider === 'static') console.log(`          model:   ${modelDir(e.model)} (${hasModel ? 'present' : `missing, ${isDownloadable(e.model) ? `fetches automatically at first search, or run \\`${ctx.name} download\\` to prefetch` : `no ${MODEL_FILENAMES} in that directory`}`})`);\n      if (declaredLanguages) console.log(`          languages: ${declaredLanguages.join(', ')} (declared by the model card)`);\n      // Cohere defaults to api.cohere.com with no url in the config; show it anyway so the\n      // reachability probe below has something to name.\n      if (e.url || reachable !== null) {\n        const shown = e.url ? redactUrl(e.url) : 'https://api.cohere.com (default)';\n        console.log(`          url:     ${shown}${reachable === null ? '' : reachable ? ' (reachable)' : ' (unreachable)'}`);\n      }\n      if (keyEnv) console.log(`          key:     env ${keyEnv.name} (${keyEnv.set ? 'set' : 'NOT SET'})`);\n      if (!hasModel) console.log('          vectors: off until the model is present; search runs on words and links');\n      else if (vectors) console.log(`          vectors: ${vectors.embedded} embedded, ${vectors.pending} pending (embedded on the first semantic search)`);\n      if (detectedLanguages) {\n        const total = Object.values(detectedLanguages).reduce((a, b) => a + b, 0);\n        const breakdown = Object.entries(detectedLanguages)\n          .sort((a, b) => b[1] - a[1])\n          .map(([code, n]) => `${code} ${Math.round((n / total) * 100)}%`)\n          .join(', ');\n        console.log(`          detected: ${breakdown} (${total} chunks classified)`);\n      }\n    }\n    console.log('');\n    for (const line of presetsLines(presets)) console.log(line);\n    console.log('');\n    console.log(`queries:  ${result.queries} saved (${ctx.name} --list)`);\n    const pid = result.watcherPid ? `pid ${result.watcherPid}, ` : '';\n    console.log(result.watcherHeartbeatSecondsAgo === null ? 'watcher:  none' : `watcher:  ${pid}last heartbeat ${result.watcherHeartbeatSecondsAgo}s ago`);\n    console.log(`sqlite:   busy_timeout ${result.busyTimeoutMs}ms (derived: 3x the largest reconcile this cache has recorded, floored at 30000ms)`);\n  }\n  db.close();\n};\nexport default status;\n"],"names":["dirname","presetCoverage","embedConfig","featureStates","SUPPORTED_CONFIG_VERSION","docCount","getMeta","open","SCHEMA_VERSION","probeReachable","languageDistribution","isDownloadable","MODEL_FILENAMES","modelDir","modelPresent","readLanguages","featuresLine","presetsLines","USAGE","CONFIG","FORMAT","formatOf","parse","printWarnings","vectorState","db","row","prepare","get","parseErrorCount","n","redactUrl","url","parsed","URL","username","password","toString","replace","embedReachable","e","provider","base","status","ctx","process","cfg","values","argv","name","format","resolveConfig","config","dbPath","warnings","features","hasModel","reachable","vectors","presets","staticLanguages","model","undefined","declaredLanguages","length","detectedLanguages","heartbeat","busyTimeoutMs","timeout","keyEnv","key","set","env","result","configPath","configVersion","migratedFrom","unknownConfigKeys","unknownKeys","tree","baseDir","cache","cacheSchema","docs","unparseableFrontmatter","on","featuresOff","off","embed","dir","downloaded","languages","queries","Object","keys","watcherPid","watcherHeartbeatSecondsAgo","Math","round","Date","now","console","log","JSON","stringify","migrated","join","shown","embedded","pending","total","reduce","a","b","breakdown","entries","sort","map","code","line","pid","close"],"mappings":"AAAA,SAASA,OAAO,QAAQ,YAAY;AACpC,SAASC,cAAc,QAAQ,uBAAuB;AACtD,SAASC,WAAW,EAAEC,aAAa,EAAEC,wBAAwB,QAAQ,qBAAqB;AAC1F,SAASC,QAAQ,EAAEC,OAAO,EAAEC,IAAI,EAAEC,cAAc,QAAQ,iBAAiB;AACzE,SAASC,cAAc,QAAQ,mBAAmB;AAClD,SAASC,oBAAoB,QAAQ,sBAAsB;AAC3D,SAASC,cAAc,EAAEC,eAAe,EAAEC,QAAQ,EAAEC,YAAY,EAAEC,aAAa,QAAQ,oBAAoB;AAC3G,SAASC,YAAY,EAAEC,YAAY,QAAQ,eAAe;AAC1D,SAASC,KAAK,QAAQ,aAAa;AACnC,SAASC,MAAM,EAAEC,MAAM,EAAEC,QAAQ,EAAEC,KAAK,EAAEC,aAAa,QAAQ,cAAc;AAG7E,0FAA0F;AAC1F,+FAA+F;AAC/F,4FAA4F;AAC5F,kGAAkG;AAElG,2FAA2F;AAC3F,4FAA4F;AAC5F,yDAAyD;AACzD,SAASC,YAAYC,EAAiC;IACpD,IAAI;QACF,MAAMC,MAAMD,GAAGE,OAAO,CAAC,yFAAyFC,GAAG;QACnH,OAAOF;IACT,EAAE,OAAM;QACN,OAAO;IACT;AACF;AAEA,SAASG,gBAAgBJ,EAAiC;IACxD,MAAMC,MAAMD,GAAGE,OAAO,CAAC,CAAC,sEAAsE,CAAC,EAAEC,GAAG;IACpG,OAAOF,IAAII,CAAC;AACd;AAEA,8FAA8F;AAC9F,qEAAqE;AACrE,OAAO,SAASC,UAAUC,GAAW;IACnC,IAAI;QACF,MAAMC,SAAS,IAAIC,IAAIF;QACvB,IAAI,CAACC,OAAOE,QAAQ,IAAI,CAACF,OAAOG,QAAQ,EAAE,OAAOJ;QACjDC,OAAOE,QAAQ,GAAG;QAClBF,OAAOG,QAAQ,GAAG;QAClB,OAAO,GAAGH,OAAOI,QAAQ,GAAGC,OAAO,CAAC,OAAO,IAAI,uBAAuB,CAAC;IACzE,EAAE,OAAM;QACN,OAAON;IACT;AACF;AAEA,gGAAgG;AAChG,+FAA+F;AAC/F,eAAeO,eAAeC,CAAqC;QAEnDA;IADd,IAAIA,EAAEC,QAAQ,KAAK,YAAaD,EAAEC,QAAQ,KAAK,YAAY,CAACD,EAAER,GAAG,EAAG,OAAO;IAC3E,MAAMU,OAAO,EAACF,SAAAA,EAAER,GAAG,cAALQ,oBAAAA,SAAS,0BAA0BF,OAAO,CAAC,QAAQ;IACjE,OAAO7B,eAAe+B,EAAEC,QAAQ,KAAK,WAAW,GAAGC,KAAK,OAAO,CAAC,GAAGA;AACrE;AAEA,MAAMC,SAAkB,OAAOC;QAeHlC,uBAOmBmC,oBAI7BC,mBACKA,kBAWEA;IArCvB,MAAM,EAAEC,MAAM,EAAE,GAAGzB,MAAMsB,IAAII,IAAI,EAAE,CAAC,OAAO,EAAEJ,IAAIK,IAAI,CAAC,CAAC,EAAE/B,MAAMyB,MAAM,EAAE,EAAE;QAAE,GAAGvB,MAAM;QAAE,GAAGD,MAAM;IAAC;IAChG,MAAM+B,SAAS7B,SAAS0B;IACxB,MAAMD,MAAMF,IAAIO,aAAa,CAACJ,OAAOK,MAAM;IAC3C,MAAM,EAAE3B,EAAE,EAAE4B,MAAM,EAAEC,QAAQ,EAAE,GAAG/C,KAAKuC;IACtCvB,cAAc+B;IAEd,MAAMC,WAAWpD,cAAc2C;IAC/B,MAAMN,IAAItC,YAAY4C;IACtB,MAAMU,WAAWhB,IAAI1B,aAAagC,OAAO;IACzC,MAAMW,YAAYjB,IAAI,MAAMD,eAAeC,KAAK;IAChD,MAAMkB,UAAUlB,IAAIhB,YAAYC,MAAM;IACtC,MAAMkC,UAAU1D,eAAewB,IAAIqB;IACnC,MAAMc,kBAAkBpB,KAAKA,EAAEC,QAAQ,KAAK,WAAW1B,cAAcyB,EAAEqB,KAAK,IAAIC;IAChF,MAAMC,oBAAoBH,mBAAmBA,gBAAgBI,MAAM,GAAG,IAAIJ,kBAAkB;IAC5F,MAAMK,qBAAoBvD,wBAAAA,qBAAqBe,iBAArBf,mCAAAA,wBAA4B;IACtD,MAAMwD,YAAY5D,QAAQmB,IAAI;IAC9B,qFAAqF;IACrF,+EAA+E;IAC/E,oEAAoE;IACpE,MAAM0C,gBAAgB,AAAC1C,GAAGE,OAAO,CAAC,uBAAuBC,GAAG,GAA2BwC,OAAO;IAC9F,sFAAsF;IACtF,MAAMC,SAAS7B,CAAAA,cAAAA,wBAAAA,EAAG8B,GAAG,IAAG;QAAErB,MAAMT,EAAE8B,GAAG;QAAEC,KAAK,EAAC1B,qBAAAA,QAAQ2B,GAAG,CAAChC,EAAE8B,GAAG,CAAC,cAAlBzB,gCAAAA,qBAAsB,QAAQ;IAAG,IAAI;IAClF,MAAM4B,SAAS;QACbrB,QAAQN,IAAI4B,UAAU;QACtBC,eAAevE;QACfwE,YAAY,GAAE9B,oBAAAA,IAAI8B,YAAY,cAAhB9B,+BAAAA,oBAAoB;QAClC+B,iBAAiB,GAAE/B,mBAAAA,IAAIgC,WAAW,cAAfhC,8BAAAA,mBAAmB;QACtCiC,MAAMjC,IAAIkC,OAAO;QACjBC,OAAOjF,QAAQqD;QACf5B,IAAI4B;QACJ6B,aAAa1E;QACb2E,MAAM9E,SAASoB;QACf2D,wBAAwBvD,gBAAgBJ;QACxC8B,UAAUA,SAAS8B,EAAE;QACrBC,aAAa/B,SAASgC,GAAG;QACzBC,OAAOhD,IAAI;YAAEC,UAAUD,EAAEC,QAAQ;YAAEoB,OAAOrB,EAAEqB,KAAK;YAAE4B,KAAKjD,EAAEC,QAAQ,KAAK,WAAW5B,SAAS2B,EAAEqB,KAAK,IAAI;YAAM7B,KAAKQ,EAAER,GAAG,GAAGD,UAAUS,EAAER,GAAG,IAAI;YAAMyB;YAAWY;YAAQqB,YAAYlC;YAAUmC,WAAW5B;YAAmBE;eAAuBP,oBAAAA,qBAAAA,UAAW,CAAC,CAAhB;QAAmB,IAAI;QACnQC;QACAiC,SAASC,OAAOC,IAAI,EAAChD,eAAAA,IAAI8C,OAAO,cAAX9C,0BAAAA,eAAe,CAAC,GAAGkB,MAAM;QAC9C+B,YAAYzF,QAAQmB,IAAI;QACxBuE,4BAA4B9B,YAAY+B,KAAKC,KAAK,CAAC,AAACC,CAAAA,KAAKC,GAAG,KAAKD,KAAK7E,KAAK,CAAC4C,UAAS,IAAK,QAAQ;QAClGC;IACF;IAEA,IAAIjB,WAAW,QAAQ;QACrBmD,QAAQC,GAAG,CAACC,KAAKC,SAAS,CAAC/B,QAAQ,MAAM;IAC3C,OAAO;QACL,MAAMgC,WAAW3D,IAAI8B,YAAY,KAAKd,YAAY,CAAC,iBAAiB,EAAEhB,IAAI8B,YAAY,CAAC,YAAY,CAAC,GAAG;QACvGyB,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE7B,OAAOrB,MAAM,CAAC,GAAG,EAAEhD,2BAA2BqG,SAAS,CAAC,CAAC;QAClF,8EAA8E;QAC9E,IAAI3D,IAAIgC,WAAW,EAAEuB,QAAQC,GAAG,CAAC,CAAC,iCAAiC,EAAExD,IAAIgC,WAAW,CAAC4B,IAAI,CAAC,OAAO;QACjGL,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE7B,OAAOM,IAAI,EAAE;QACtCsB,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE7B,OAAOQ,KAAK,CAAC,SAAS,EAAEzE,eAAe,iCAAiC,CAAC;QAClG6F,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE7B,OAAOU,IAAI,GAAGV,OAAOW,sBAAsB,GAAG,IAAI,CAAC,GAAG,EAAEX,OAAOW,sBAAsB,CAAC,8DAA8D,CAAC,GAAG,IAAI;QACrLiB,QAAQC,GAAG,CAAC;QACZD,QAAQC,GAAG,CAACtF,aAAauC;QACzB,4FAA4F;QAC5F,0BAA0B;QAC1B,IAAI,CAACf,GAAG6D,QAAQC,GAAG,CAAC;aACf;YACHD,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE9D,EAAEC,QAAQ,CAAC,CAAC,EAAED,EAAEqB,KAAK,EAAE;YAChD,IAAIrB,EAAEC,QAAQ,KAAK,UAAU4D,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAEzF,SAAS2B,EAAEqB,KAAK,EAAE,EAAE,EAAEL,WAAW,YAAY,CAAC,SAAS,EAAE7C,eAAe6B,EAAEqB,KAAK,IAAI,CAAC,gDAAgD,EAAEjB,IAAIK,IAAI,CAAC,uBAAuB,CAAC,GAAG,CAAC,GAAG,EAAErC,gBAAgB,kBAAkB,CAAC,EAAE,CAAC,CAAC,CAAC;YACvR,IAAImD,mBAAmBsC,QAAQC,GAAG,CAAC,CAAC,qBAAqB,EAAEvC,kBAAkB2C,IAAI,CAAC,MAAM,6BAA6B,CAAC;YACtH,qFAAqF;YACrF,kDAAkD;YAClD,IAAIlE,EAAER,GAAG,IAAIyB,cAAc,MAAM;gBAC/B,MAAMkD,QAAQnE,EAAER,GAAG,GAAGD,UAAUS,EAAER,GAAG,IAAI;gBACzCqE,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAEK,QAAQlD,cAAc,OAAO,KAAKA,YAAY,iBAAiB,kBAAkB;YACrH;YACA,IAAIY,QAAQgC,QAAQC,GAAG,CAAC,CAAC,uBAAuB,EAAEjC,OAAOpB,IAAI,CAAC,EAAE,EAAEoB,OAAOE,GAAG,GAAG,QAAQ,UAAU,CAAC,CAAC;YACnG,IAAI,CAACf,UAAU6C,QAAQC,GAAG,CAAC;iBACtB,IAAI5C,SAAS2C,QAAQC,GAAG,CAAC,CAAC,mBAAmB,EAAE5C,QAAQkD,QAAQ,CAAC,WAAW,EAAElD,QAAQmD,OAAO,CAAC,gDAAgD,CAAC;YACnJ,IAAI5C,mBAAmB;gBACrB,MAAM6C,QAAQjB,OAAO9C,MAAM,CAACkB,mBAAmB8C,MAAM,CAAC,CAACC,GAAGC,IAAMD,IAAIC,GAAG;gBACvE,MAAMC,YAAYrB,OAAOsB,OAAO,CAAClD,mBAC9BmD,IAAI,CAAC,CAACJ,GAAGC,IAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE,EAC1BK,GAAG,CAAC,CAAC,CAACC,MAAMxF,EAAE,GAAK,GAAGwF,KAAK,CAAC,EAAErB,KAAKC,KAAK,CAAC,AAACpE,IAAIgF,QAAS,KAAK,CAAC,CAAC,EAC9DJ,IAAI,CAAC;gBACRL,QAAQC,GAAG,CAAC,CAAC,oBAAoB,EAAEY,UAAU,EAAE,EAAEJ,MAAM,mBAAmB,CAAC;YAC7E;QACF;QACAT,QAAQC,GAAG,CAAC;QACZ,KAAK,MAAMiB,QAAQtG,aAAa0C,SAAU0C,QAAQC,GAAG,CAACiB;QACtDlB,QAAQC,GAAG,CAAC;QACZD,QAAQC,GAAG,CAAC,CAAC,UAAU,EAAE7B,OAAOmB,OAAO,CAAC,QAAQ,EAAEhD,IAAIK,IAAI,CAAC,QAAQ,CAAC;QACpE,MAAMuE,MAAM/C,OAAOsB,UAAU,GAAG,CAAC,IAAI,EAAEtB,OAAOsB,UAAU,CAAC,EAAE,CAAC,GAAG;QAC/DM,QAAQC,GAAG,CAAC7B,OAAOuB,0BAA0B,KAAK,OAAO,mBAAmB,CAAC,UAAU,EAAEwB,IAAI,eAAe,EAAE/C,OAAOuB,0BAA0B,CAAC,KAAK,CAAC;QACtJK,QAAQC,GAAG,CAAC,CAAC,uBAAuB,EAAE7B,OAAON,aAAa,CAAC,kFAAkF,CAAC;IAChJ;IACA1C,GAAGgG,KAAK;AACV;AACA,eAAe9E,OAAO"}

package/dist/esm/commands/related.js

@@ -1,6 +1,7 @@
-import { embedEnabled, resolveSearch } from '../config/index.js';
+import { embedConfig, embedEnabled, resolveSearch } from '../config/index.js';
+import { embedPending, hasEmbedding, similarNotes } from '../embed/query.js';
+import { localModelMissing, MODEL_FILENAMES } from '../embed/store.js';
 import { SenseError } from '../errors.js';
-import { embedPending, hasEmbedding, modelPresent, similarNotes } from '../features/embed.js';
 import { resolveNote } from './peek.js';
 import { scopedPaths, scopeHasEmbeddings } from './scope.js';
 // Notes most similar by cosine, excluding self and everything already linked either way.
@@ -20,16 +21,19 @@ export async function relatedNotes(db, cfg, pathArg, overrides, k) {
     // note does not already link to, which is a real answer.
     const effective = resolveSearch(cfg, overrides);
     if (!embedEnabled(cfg)) {
-        throw new SenseError('EMBED_DISABLED', 'related ranks notes by meaning, and this tree has no embedding model; add an "embed" block naming one to sense.config.json, then run `sense download` (search works without it, on words and links)');
+        throw new SenseError('EMBED_DISABLED', 'related ranks notes by meaning, and this tree has no embedding model; add an "embed" block naming one to sense.config.json (a Hugging Face id fetches automatically at first use; `sense download` prefetches it) -- search works without it, on words and links');
     }
-    // search gates on the same flag (see wantsVectors above); reading it here too keeps
-    // `semantic: false` meaning one thing. Without this, an overlapping semantic-on preset's
-    // vectors would answer for a scope that declined them.
-    if (!effective.semantic) {
-        throw new SenseError('PRESET_NOT_SEMANTIC', `preset "${effective.presetName}" sets "semantic": false, so this scope has no vectors and related has no other signal; search it instead (words and links), or set semantic back on for that preset`);
+    // search gates on the same signal (see wantsVectors above); reading it here too keeps a
+    // preset's declared signals meaning one thing. Without this, an overlapping vectors-on
+    // preset's vectors would answer for a scope that declined them.
+    if (effective.signals.vectors === undefined) {
+        throw new SenseError('PRESET_NOT_SEMANTIC', `preset "${effective.presetName}" has no "vectors" signal, so this scope has no vectors and related has no other signal; search it instead (words and links), or add "vectors" to that preset's signals`);
     }
-    if (!modelPresent(cfg)) {
-        throw new SenseError('EMBED_MODEL_MISSING', 'related ranks notes by meaning, so it needs the embedding model, which is not downloaded; run `sense download` (search still works without it, on words and links)');
+    // A downloadable HF id proceeds -- embedPending's getProvider call fetches it lazily on
+    // consent. Only a local path with missing files can never fetch itself.
+    const e = embedConfig(cfg); // embedEnabled(cfg) above guarantees this is set
+    if (localModelMissing(e)) {
+        throw new SenseError('EMBED_MODEL_MISSING', `related ranks notes by meaning, so it needs the embedding model, but the local model path "${e.model}" is missing ${MODEL_FILENAMES}; point embed.model at a directory containing them (search still works without it, on words and links)`);
     }
     const allowed = scopedPaths(db, cfg, overrides);
     // Top up pending rows before the seed check, or a fresh index reports every note as

package/dist/esm/commands/related.js.map

@@ -1 +1 @@
-{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/related.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { embedEnabled, resolveSearch } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { embedPending, hasEmbedding, modelPresent, similarNotes } from '../features/embed.ts';\nimport { resolveNote } from './peek.ts';\nimport { scopedPaths, scopeHasEmbeddings } from './scope.ts';\n\n// Notes most similar by cosine, excluding self and everything already linked either way.\n// Its own command, not a peek section: a full embeddings scan is ~480ms at 26k notes.\nexport async function relatedNotes(db: DatabaseSync, cfg: ResolvedConfig, pathArg: string, overrides: SearchOverrides, k: number): Promise<Array<{ path: string; similarity: number }>> {\n  const paths = (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n  const path = resolveNote(paths, pathArg);\n\n  const outbound = (db.prepare('SELECT DISTINCT dst FROM links WHERE src = ? AND dst IS NOT NULL AND dst != src').all(path) as Array<{ dst: string }>).map((r) => r.dst);\n  const backlinks = (db.prepare('SELECT DISTINCT src FROM links WHERE dst = ? AND src != dst').all(path) as Array<{ src: string }>).map((r) => r.src);\n  const exclude = new Set([path, ...outbound, ...backlinks]);\n\n  // Vectors are the only signal `related` has, so every way of not having them is an error\n  // naming the cause. An empty table then means one thing: nothing near in meaning that this\n  // note does not already link to, which is a real answer.\n  const effective = resolveSearch(cfg, overrides);\n  if (!embedEnabled(cfg)) {\n    throw new SenseError('EMBED_DISABLED', 'related ranks notes by meaning, and this tree has no embedding model; add an \"embed\" block naming one to sense.config.json, then run `sense download` (search works without it, on words and links)');\n  }\n  // search gates on the same flag (see wantsVectors above); reading it here too keeps\n  // `semantic: false` meaning one thing. Without this, an overlapping semantic-on preset's\n  // vectors would answer for a scope that declined them.\n  if (!effective.semantic) {\n    throw new SenseError('PRESET_NOT_SEMANTIC', `preset \"${effective.presetName}\" sets \"semantic\": false, so this scope has no vectors and related has no other signal; search it instead (words and links), or set semantic back on for that preset`);\n  }\n  if (!modelPresent(cfg)) {\n    throw new SenseError('EMBED_MODEL_MISSING', 'related ranks notes by meaning, so it needs the embedding model, which is not downloaded; run `sense download` (search still works without it, on words and links)');\n  }\n  const allowed = scopedPaths(db, cfg, overrides);\n  // Top up pending rows before the seed check, or a fresh index reports every note as\n  // having no indexed text until some search has run.\n  await embedPending(db, cfg, cfg.baseDir);\n  if (!hasEmbedding(db, path)) {\n    throw new SenseError('NOTE_NOT_EMBEDDED', `${path} has no indexed text to compare -- a note that is frontmatter only, or empty, has nothing to rank by meaning`);\n  }\n  if (!scopeHasEmbeddings(db, cfg, allowed)) return [];\n  return similarNotes(db, cfg, path, { exclude, allowed, k });\n}\n"],"names":["embedEnabled","resolveSearch","SenseError","embedPending","hasEmbedding","modelPresent","similarNotes","resolveNote","scopedPaths","scopeHasEmbeddings","relatedNotes","db","cfg","pathArg","overrides","k","paths","prepare","all","map","r","path","outbound","dst","backlinks","src","exclude","Set","effective","semantic","presetName","allowed","baseDir"],"mappings":"AAEA,SAASA,YAAY,EAAEC,aAAa,QAAQ,qBAAqB;AACjE,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,YAAY,EAAEC,YAAY,EAAEC,YAAY,EAAEC,YAAY,QAAQ,uBAAuB;AAC9F,SAASC,WAAW,QAAQ,YAAY;AACxC,SAASC,WAAW,EAAEC,kBAAkB,QAAQ,aAAa;AAE7D,yFAAyF;AACzF,sFAAsF;AACtF,OAAO,eAAeC,aAAaC,EAAgB,EAAEC,GAAmB,EAAEC,OAAe,EAAEC,SAA0B,EAAEC,CAAS;IAC9H,MAAMC,QAAQ,AAACL,GAAGM,OAAO,CAAC,kCAAkCC,GAAG,GAA+BC,GAAG,CAAC,CAACC,IAAMA,EAAEC,IAAI;IAC/G,MAAMA,OAAOd,YAAYS,OAAOH;IAEhC,MAAMS,WAAW,AAACX,GAAGM,OAAO,CAAC,mFAAmFC,GAAG,CAACG,MAAiCF,GAAG,CAAC,CAACC,IAAMA,EAAEG,GAAG;IACrK,MAAMC,YAAY,AAACb,GAAGM,OAAO,CAAC,+DAA+DC,GAAG,CAACG,MAAiCF,GAAG,CAAC,CAACC,IAAMA,EAAEK,GAAG;IAClJ,MAAMC,UAAU,IAAIC,IAAI;QAACN;WAASC;WAAaE;KAAU;IAEzD,yFAAyF;IACzF,2FAA2F;IAC3F,yDAAyD;IACzD,MAAMI,YAAY3B,cAAcW,KAAKE;IACrC,IAAI,CAACd,aAAaY,MAAM;QACtB,MAAM,IAAIV,WAAW,kBAAkB;IACzC;IACA,oFAAoF;IACpF,yFAAyF;IACzF,uDAAuD;IACvD,IAAI,CAAC0B,UAAUC,QAAQ,EAAE;QACvB,MAAM,IAAI3B,WAAW,uBAAuB,CAAC,QAAQ,EAAE0B,UAAUE,UAAU,CAAC,oKAAoK,CAAC;IACnP;IACA,IAAI,CAACzB,aAAaO,MAAM;QACtB,MAAM,IAAIV,WAAW,uBAAuB;IAC9C;IACA,MAAM6B,UAAUvB,YAAYG,IAAIC,KAAKE;IACrC,oFAAoF;IACpF,oDAAoD;IACpD,MAAMX,aAAaQ,IAAIC,KAAKA,IAAIoB,OAAO;IACvC,IAAI,CAAC5B,aAAaO,IAAIU,OAAO;QAC3B,MAAM,IAAInB,WAAW,qBAAqB,GAAGmB,KAAK,4GAA4G,CAAC;IACjK;IACA,IAAI,CAACZ,mBAAmBE,IAAIC,KAAKmB,UAAU,OAAO,EAAE;IACpD,OAAOzB,aAAaK,IAAIC,KAAKS,MAAM;QAAEK;QAASK;QAAShB;IAAE;AAC3D"}
+{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/related.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { embedConfig, embedEnabled, resolveSearch } from '../config/index.ts';\nimport { embedPending, hasEmbedding, similarNotes } from '../embed/query.ts';\nimport { localModelMissing, MODEL_FILENAMES } from '../embed/store.ts';\nimport { SenseError } from '../errors.ts';\nimport { resolveNote } from './peek.ts';\nimport { scopedPaths, scopeHasEmbeddings } from './scope.ts';\n\n// Notes most similar by cosine, excluding self and everything already linked either way.\n// Its own command, not a peek section: a full embeddings scan is ~480ms at 26k notes.\nexport async function relatedNotes(db: DatabaseSync, cfg: ResolvedConfig, pathArg: string, overrides: SearchOverrides, k: number): Promise<Array<{ path: string; similarity: number }>> {\n  const paths = (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n  const path = resolveNote(paths, pathArg);\n\n  const outbound = (db.prepare('SELECT DISTINCT dst FROM links WHERE src = ? AND dst IS NOT NULL AND dst != src').all(path) as Array<{ dst: string }>).map((r) => r.dst);\n  const backlinks = (db.prepare('SELECT DISTINCT src FROM links WHERE dst = ? AND src != dst').all(path) as Array<{ src: string }>).map((r) => r.src);\n  const exclude = new Set([path, ...outbound, ...backlinks]);\n\n  // Vectors are the only signal `related` has, so every way of not having them is an error\n  // naming the cause. An empty table then means one thing: nothing near in meaning that this\n  // note does not already link to, which is a real answer.\n  const effective = resolveSearch(cfg, overrides);\n  if (!embedEnabled(cfg)) {\n    throw new SenseError('EMBED_DISABLED', 'related ranks notes by meaning, and this tree has no embedding model; add an \"embed\" block naming one to sense.config.json (a Hugging Face id fetches automatically at first use; `sense download` prefetches it) -- search works without it, on words and links');\n  }\n  // search gates on the same signal (see wantsVectors above); reading it here too keeps a\n  // preset's declared signals meaning one thing. Without this, an overlapping vectors-on\n  // preset's vectors would answer for a scope that declined them.\n  if (effective.signals.vectors === undefined) {\n    throw new SenseError('PRESET_NOT_SEMANTIC', `preset \"${effective.presetName}\" has no \"vectors\" signal, so this scope has no vectors and related has no other signal; search it instead (words and links), or add \"vectors\" to that preset's signals`);\n  }\n  // A downloadable HF id proceeds -- embedPending's getProvider call fetches it lazily on\n  // consent. Only a local path with missing files can never fetch itself.\n  const e = embedConfig(cfg); // embedEnabled(cfg) above guarantees this is set\n  if (localModelMissing(e)) {\n    throw new SenseError('EMBED_MODEL_MISSING', `related ranks notes by meaning, so it needs the embedding model, but the local model path \"${e.model}\" is missing ${MODEL_FILENAMES}; point embed.model at a directory containing them (search still works without it, on words and links)`);\n  }\n  const allowed = scopedPaths(db, cfg, overrides);\n  // Top up pending rows before the seed check, or a fresh index reports every note as\n  // having no indexed text until some search has run.\n  await embedPending(db, cfg, cfg.baseDir);\n  if (!hasEmbedding(db, path)) {\n    throw new SenseError('NOTE_NOT_EMBEDDED', `${path} has no indexed text to compare -- a note that is frontmatter only, or empty, has nothing to rank by meaning`);\n  }\n  if (!scopeHasEmbeddings(db, cfg, allowed)) return [];\n  return similarNotes(db, cfg, path, { exclude, allowed, k });\n}\n"],"names":["embedConfig","embedEnabled","resolveSearch","embedPending","hasEmbedding","similarNotes","localModelMissing","MODEL_FILENAMES","SenseError","resolveNote","scopedPaths","scopeHasEmbeddings","relatedNotes","db","cfg","pathArg","overrides","k","paths","prepare","all","map","r","path","outbound","dst","backlinks","src","exclude","Set","effective","signals","vectors","undefined","presetName","e","model","allowed","baseDir"],"mappings":"AAEA,SAASA,WAAW,EAAEC,YAAY,EAAEC,aAAa,QAAQ,qBAAqB;AAC9E,SAASC,YAAY,EAAEC,YAAY,EAAEC,YAAY,QAAQ,oBAAoB;AAC7E,SAASC,iBAAiB,EAAEC,eAAe,QAAQ,oBAAoB;AACvE,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,WAAW,QAAQ,YAAY;AACxC,SAASC,WAAW,EAAEC,kBAAkB,QAAQ,aAAa;AAE7D,yFAAyF;AACzF,sFAAsF;AACtF,OAAO,eAAeC,aAAaC,EAAgB,EAAEC,GAAmB,EAAEC,OAAe,EAAEC,SAA0B,EAAEC,CAAS;IAC9H,MAAMC,QAAQ,AAACL,GAAGM,OAAO,CAAC,kCAAkCC,GAAG,GAA+BC,GAAG,CAAC,CAACC,IAAMA,EAAEC,IAAI;IAC/G,MAAMA,OAAOd,YAAYS,OAAOH;IAEhC,MAAMS,WAAW,AAACX,GAAGM,OAAO,CAAC,mFAAmFC,GAAG,CAACG,MAAiCF,GAAG,CAAC,CAACC,IAAMA,EAAEG,GAAG;IACrK,MAAMC,YAAY,AAACb,GAAGM,OAAO,CAAC,+DAA+DC,GAAG,CAACG,MAAiCF,GAAG,CAAC,CAACC,IAAMA,EAAEK,GAAG;IAClJ,MAAMC,UAAU,IAAIC,IAAI;QAACN;WAASC;WAAaE;KAAU;IAEzD,yFAAyF;IACzF,2FAA2F;IAC3F,yDAAyD;IACzD,MAAMI,YAAY5B,cAAcY,KAAKE;IACrC,IAAI,CAACf,aAAaa,MAAM;QACtB,MAAM,IAAIN,WAAW,kBAAkB;IACzC;IACA,wFAAwF;IACxF,uFAAuF;IACvF,gEAAgE;IAChE,IAAIsB,UAAUC,OAAO,CAACC,OAAO,KAAKC,WAAW;QAC3C,MAAM,IAAIzB,WAAW,uBAAuB,CAAC,QAAQ,EAAEsB,UAAUI,UAAU,CAAC,uKAAuK,CAAC;IACtP;IACA,wFAAwF;IACxF,wEAAwE;IACxE,MAAMC,IAAInC,YAAYc,MAAM,iDAAiD;IAC7E,IAAIR,kBAAkB6B,IAAI;QACxB,MAAM,IAAI3B,WAAW,uBAAuB,CAAC,2FAA2F,EAAE2B,EAAEC,KAAK,CAAC,aAAa,EAAE7B,gBAAgB,sGAAsG,CAAC;IAC1R;IACA,MAAM8B,UAAU3B,YAAYG,IAAIC,KAAKE;IACrC,oFAAoF;IACpF,oDAAoD;IACpD,MAAMb,aAAaU,IAAIC,KAAKA,IAAIwB,OAAO;IACvC,IAAI,CAAClC,aAAaS,IAAIU,OAAO;QAC3B,MAAM,IAAIf,WAAW,qBAAqB,GAAGe,KAAK,4GAA4G,CAAC;IACjK;IACA,IAAI,CAACZ,mBAAmBE,IAAIC,KAAKuB,UAAU,OAAO,EAAE;IACpD,OAAOhC,aAAaQ,IAAIC,KAAKS,MAAM;QAAEK;QAASS;QAASpB;IAAE;AAC3D"}

package/dist/esm/commands/search.js

@@ -1,20 +1,14 @@
 import { readFileSync } from 'node:fs';
 import { join } from 'node:path';
-import { anyPresetEmbeds, contentTokenize, featureEnabled, resolveSearch } from '../config/index.js';
+import { contentTokenize, embedConfig, featureEnabled, resolveSearch } from '../config/index.js';
+import { localModelMissing, MODEL_FILENAMES } from '../embed/store.js';
 import { SenseError } from '../errors.js';
-import { modelPresent, semanticCandidates } from '../features/embed.js';
-import { linkEdges } from '../features/index.js';
-import { personalizedRank } from '../graph.js';
 import { searchError } from '../search-error.js';
 import { segmentMatch } from '../segment.js';
 import { materializeScope, narrowByWhere, rawScope, scopeHasEmbeddings } from './scope.js';
-// Mirrors the main columns onto the sidecars, so a title match found through title_seg ranks
-// like a title match found through title, not like a body match.
-const WEIGHTED_BM25 = 'bm25(content, 10.0, 5.0, 1.0, 0, 10.0, 5.0, 1.0)';
-const RRF_K = 60;
+import { linksCandidates, vectorsCandidates, wordsCandidates } from './signals.js';
 // snippet() re-tokenizes each candidate doc, superlinearly: ~10s for one 1MB doc
 // (BENCHMARKING.md). Past this bound, rows get the linear JS excerpt instead.
-const SNIPPET_BOUND = 16384;
 const EXCERPT_WINDOW = 160;
 // Bare terms from an FTS5 query string: strips operators/quoting so the oversized-doc excerpt
 // scan matches the same words the query matched on, not FTS5 syntax.
@@ -107,12 +101,12 @@ function lineRangeFor(db, path, line) {
     const row = db.prepare('SELECT start_line, end_line FROM sections WHERE "path" = ? AND start_line <= ? AND end_line >= ? ORDER BY start_line DESC LIMIT 1').get(path, line, line);
     return row ? `L${row.start_line}-${row.end_line}` : null;
 }
-// BM25 + link expansion + vectors, fused by reciprocal rank; `via` names the signal per row.
+// The declared or defaulted signals compose via RRF; `via` names which ones produced each row.
 // `opts` arrives already resolved (config.ts:resolveSearch).
 export async function search(db, cfg, terms, opts = {}) {
     var _hits_get, _chunkLines_get, _chunkSimilarity_get;
     const effective = resolveSearch(cfg, opts);
-    const { k } = effective;
+    const { k, signals } = effective;
     const allPaths = db.prepare('SELECT "path" FROM frontmatter').all().map((r)=>r.path);
     const scopePaths = rawScope(db, cfg, opts, allPaths);
     const scopeActive = scopePaths.size < allPaths.length;
@@ -126,11 +120,14 @@ export async function search(db, cfg, terms, opts = {}) {
         throw searchError(err, terms, effective.where);
     }
     const fetch = Math.max(k * 3, 30);
-    // Asking for vectors without a model is a misconfiguration, not a mode: degrading would make
-    // the same search answer differently before and after a download. semantic:false never asks.
-    const wantsVectors = effective.semantic && anyPresetEmbeds(cfg);
-    if (wantsVectors && !modelPresent(cfg)) {
-        throw new SenseError('EMBED_MODEL_MISSING', `preset "${effective.presetName}" searches with vectors, but the embedding model is not available; run \`sense download\`, or set "semantic": false on that preset to search on words and links`);
+    // A downloadable HF id proceeds -- getProvider fetches it lazily on consent. Only a
+    // local path with missing files errors here, since nothing will ever fetch it for itself.
+    const wantsVectors = signals.vectors !== undefined;
+    if (wantsVectors) {
+        const e = embedConfig(cfg); // validate.ts guarantees this is set whenever "vectors" is declared
+        if (localModelMissing(e)) {
+            throw new SenseError('EMBED_MODEL_MISSING', `preset "${effective.presetName}" searches with vectors, but the local model path "${e.model}" is missing ${MODEL_FILENAMES}; point embed.model at a directory containing them, or drop "vectors" from that preset's signals to search without them`);
+        }
     }
     const semanticEnabled = wantsVectors && scopeHasEmbeddings(db, cfg, allowedPaths);
     // Terms pass to MATCH as written, with one transform: a run of text whose language marks no
@@ -152,18 +149,17 @@ export async function search(db, cfg, terms, opts = {}) {
     // thousands of paths, past SQLITE_MAX_VARIABLE_NUMBER on older builds.
     if (scopeActive) materializeScope(db, '_search_scope', scopePaths);
     const scopeCond = scopeActive ? `AND content.path IN (SELECT "path" FROM _search_scope)` : '';
-    // Docs past SNIPPET_BOUND skip snippet() entirely -- SQLite
-    // short-circuits the untaken CASE branch, so it's never invoked on them.
-    // Column 2 (text), not -1 (best column): -1 could surface a _seg sidecar as the excerpt,
-    // which is machine-spaced and not what the author wrote. A row that matches only in a
-    // sidecar gets an unhighlighted text excerpt below instead -- a stated cost.
-    const matchSql = `SELECT content.path AS path, CASE WHEN length(content.text) <= ${SNIPPET_BOUND} THEN snippet(content, 2, '«', '»', '…', 10) ELSE NULL END AS hit FROM content ${whereJoin} WHERE content MATCH ? ${whereCond} ${scopeCond} ORDER BY ${WEIGHTED_BM25} LIMIT ${fetch}`;
     const query = contentTokenize(cfg) === undefined ? segmentMatch(terms) : terms;
-    let matchRows;
-    try {
-        matchRows = db.prepare(matchSql).all(query);
-    } catch (err) {
-        throw searchError(err, terms, scope);
+    const candidates = new Map();
+    let matchRows = [];
+    // A query that rewrites to nothing (e.g. only unspaced-script punctuation) has no
+    // searchable words: zero lexical rows, and the other signals still compose.
+    if (signals.words !== undefined && query.trim() !== '') {
+        try {
+            matchRows = wordsCandidates(db, candidates, query, whereJoin, whereCond, scopeCond, fetch, signals.words);
+        } catch (err) {
+            throw searchError(err, terms, scope);
+        }
     }
     const hits = new Map(matchRows.map((r)=>[
             r.path,
@@ -172,58 +168,11 @@ export async function search(db, cfg, terms, opts = {}) {
     // hit === null here means the bound suppressed snippet(), not "no match" -- distinguish
     // from via='link' rows (never in matchRows, so absent from this set) below.
     const oversized = new Set(matchRows.filter((r)=>r.hit === null).map((r)=>r.path));
-    const candidates = new Map();
-    matchRows.forEach((r, i)=>{
-        candidates.set(r.path, {
-            score: 1 / (RRF_K + i),
-            via: 'match'
-        });
-    });
-    const edges = featureEnabled(cfg, 'links') && matchRows.length > 0 ? linkEdges(db) : [];
-    if (edges.length > 0) {
-        // Gates the label only, not the score: restart mass ranks every seed without an incident
-        // edge, but dropping it from the score reweights fusion (FEVER hit@10 0.997 -> 0.907).
-        const linked = new Set(edges.flat());
-        const seeds = new Map(matchRows.map((r, i)=>[
-                r.path,
-                1 / (i + 1)
-            ]));
-        const ranked = [
-            ...personalizedRank(allPaths, edges, seeds)
-        ].filter(([path, score])=>score > 1e-9 && allowedPaths.has(path)).sort((a, b)=>b[1] - a[1]).slice(0, fetch);
-        ranked.forEach(([path], i)=>{
-            const existing = candidates.get(path);
-            if (existing) {
-                existing.score += 1 / (RRF_K + i);
-                if (linked.has(path)) existing.via = 'match+link';
-            } else {
-                candidates.set(path, {
-                    score: 1 / (RRF_K + i),
-                    via: 'link'
-                });
-            }
-        });
-    }
-    // Vector expansion, invoked only: a third RRF list at the swept flat-region constants
-    // (weight 1, pool = fetch). Each row carries its best chunk's line range.
-    const chunkLines = new Map();
-    const chunkSimilarity = new Map();
+    if (signals.links !== undefined) linksCandidates(db, candidates, matchRows, allPaths, allowedPaths, fetch, signals.links);
+    let chunkLines = new Map();
+    let chunkSimilarity = new Map();
     if (semanticEnabled) {
-        const vec = await semanticCandidates(db, cfg, terms, fetch, allowedPaths);
-        vec.forEach(({ path, lines, similarity }, i)=>{
-            chunkLines.set(path, lines);
-            chunkSimilarity.set(path, similarity);
-            const existing = candidates.get(path);
-            if (existing) {
-                existing.score += 1 / (RRF_K + i);
-                existing.via = `${existing.via}+vector`;
-            } else {
-                candidates.set(path, {
-                    score: 1 / (RRF_K + i),
-                    via: 'vector'
-                });
-            }
-        });
+        ({ chunkLines, chunkSimilarity } = await vectorsCandidates(db, cfg, candidates, terms, fetch, allowedPaths, signals.vectors));
     }
     db.exec('DROP TABLE IF EXISTS _search');
     db.exec('CREATE TEMP TABLE _search ("path" TEXT PRIMARY KEY, score REAL, via TEXT, hit TEXT, lines TEXT, similarity REAL)');