Update

adapters.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eaf4e948407ee24dc86108eaeceae31e17082d7d66bd626338ebfe8fd8a6ddba
+oid sha256:210aef249494e846fa2a1d0810260dafca7c618d273f444433b1389acda5ae2a
 size 67323724

manifest.json

@@ -1,20 +1,20 @@
 {
-  "id": "nano-planner-modality-v2",
+  "id": "nano-simple-questions-step-293",
   "base_repo": "thoddnn/Qwen3.5-2B-4bit",
   "base_revision": "da5c842efe38a8404c33e02bc9c59b2d81c99c34",
-  "schema": "nano-planner-modality-v2",
+  "schema": "nano-planner-modality-v3",
   "files": {
     "adapter_config.json": {
-      "sha256": "4116ad764b4a283dd3c418c5b3e3e3862b86e96dfcb6f69c27136890cd910766",
-      "bytes": 10880
+      "bytes": 10880,
+      "sha256": "4116ad764b4a283dd3c418c5b3e3e3862b86e96dfcb6f69c27136890cd910766"
     },
     "adapters.safetensors": {
-      "sha256": "eaf4e948407ee24dc86108eaeceae31e17082d7d66bd626338ebfe8fd8a6ddba",
-      "bytes": 67323724
+      "bytes": 67323724,
+      "sha256": "210aef249494e846fa2a1d0810260dafca7c618d273f444433b1389acda5ae2a"
     },
     "prompt.txt": {
-      "sha256": "eecbf32a4682dbdf1099fdbc571ba9b0aba61870ff559e19904ebbcf369fc62b",
-      "bytes": 8592
+      "bytes": 10522,
+      "sha256": "5927a2ebd42d74d964999ec191b87ee7ba3038bb4d7992a75a856baf37a5d132"
     }
   }
 }

prompt.txt

@@ -1,10 +1,10 @@
 Plan a local-file search. Return only JSON:
 {"language":"en","file_types":[],"searches":[{"query":"retrieval terms","mode":"keyword"}],"facts":[{"question":"Complete question?","scope":"any_unit","type":"boolean","modality":"hybrid"}],"match_mode":"first"}.
 Use the user's language for searches and questions; preserve names and quoted text. Keys and enum values stay English. Preserve ALL requirements without inventing any.
-file_types: only explicitly requested formats. Categories: image,pdf,video,audio,document,word,spreadsheet,presentation,email,code,ebook. Or an explicitly named extension without the dot: 3fr,aac,ai,aiff,arw,avi,avif,bash,bat,bear,bmp,c,cfg,cmd,conf,cpp,cr2,cr3,craft,crw,css,csv,dcr,dng,doc,docx,eml,emlx,epub,fig,flac,gif,go,h,heic,heif,hpp,html,indd,ini,ipynb,java,jpeg,jpg,js,json,jsx,key,kt,less,log,m4a,m4v,mbox,md,mef,mkv,mov,mp3,mp4,mrw,mts,mxf,nef,notes,notion,nrw,odp,ods,odt,ogg,olk15msgsource,orf,pages,pdf,pef,php,png,ppt,pptx,ps1,psd,ptx,py,r,raf,rb,rst,rtf,rw2,sass,scala,scss,sh,sketch,sql,sr2,srw,swift,tif,tiff,toml,ts,tsx,txt,wav,webarchive,webp,wmv,xls,xlsx,xml,yaml. Otherwise []. Use audio for audio files, email for emails of any supported format; do not narrow a category to one extension unless requested. Formats are metadata filters, never verification facts. Do not assume invoices are PDFs.
+file_types: only explicitly requested formats. Categories: image,pdf,video,audio,document,word,spreadsheet,presentation,email,code,ebook. Or an explicitly named extension without the dot: 3fr,aac,ai,aiff,arw,avi,avif,bash,bat,bear,bmp,c,cfg,cmd,conf,cpp,cr2,cr3,craft,crw,css,csv,dcr,dng,doc,docx,eml,emlx,epub,fig,flac,gif,go,h,heic,heif,hpp,html,indd,ini,ipynb,java,jpeg,jpg,js,json,jsx,key,kt,less,log,m4a,m4v,mbox,md,mef,mkv,mov,mp3,mp4,mrw,mts,mxf,nef,notes,notion,nrw,odp,ods,odt,ogg,olk15msgsource,orf,pages,pdf,pef,php,png,ppt,pptx,ps1,psd,ptx,py,r,raf,rb,rst,rtf,rw2,sass,scala,scss,sh,sketch,sql,sr2,srw,swift,tif,tiff,toml,ts,tsx,txt,wav,webarchive,webp,wmv,xls,xlsx,xml,yaml. Otherwise []. Use audio for audio files, email for emails of any supported format; do not narrow a category to one extension unless requested. Map requested application names to their native extensions (Figma: fig; Excel: xls,xlsx; Word: doc,docx; PowerPoint: ppt,pptx; Photoshop: psd). A specific extension narrows its category: image in .cr3 format means ["cr3"], not ["image","cr3"]. Keep explicit alternatives such as PDF or DOCX. Formats are metadata filters, never verification facts. Do not assume invoices are PDFs.
 For an explicitly requested recording with neither medium specified, allow both audio and video categories; filters are alternatives, not simultaneous requirements.
-searches: 1-3 useful queries in execution order. Focus first, broaden when useful. Use semantic for visual subjects and conceptual meaning; keyword for document terms/names; exact for literal text; filename only for explicitly named files. For quoted text, keyword AND exact searches with the same text can help. Never duplicate the same query AND mode. Keep comparisons, scope, page/row constraints and output instructions out of retrieval terms. Meaningful identifiers, event dates and visual counts may help retrieval. Search distinct visual subjects separately when useful. Automatic keyword/semantic fallback already exists. Do not pad the list.
-facts: 1-8 questions; all must pass. Boolean asks a question whose YES satisfies the requirement; omit target. Use boolean for comparisons, ranges, negation, identity, relations, OR alternatives and substring containment. Exact values use year, count or value with a short string target. Do not generate answer options. Preserve strict/inclusive bounds, units, party roles and correct date/amount fields: payslip year means pay period; final total differs from balance due and individual line amounts. About a business does not mean issued by it.
+searches: 1-3 useful queries in execution order. Focus first, broaden when useful. For image or visual-content requests use semantic ONLY, including visible text, signatures, diagrams, rendered pages and attached images. Never use keyword, exact or filename for these requests; retain explicit filename requirements as facts. Otherwise use semantic for conceptual meaning; keyword for document terms/names; exact for literal text; filename only for explicitly named files. For quoted nonvisual text, keyword AND exact searches with the same text can help. Never duplicate the same query AND mode. Keep comparisons, scope, page/row constraints and output instructions out of retrieval terms. Meaningful identifiers, event dates and visual counts may help retrieval. Search distinct visual subjects separately when useful. Automatic keyword/semantic fallback already exists. Do not pad the list.
+facts: 0-8 questions; all generated facts must pass. Boolean asks a question whose YES satisfies the requirement; omit target. Use boolean for comparisons, ranges, negation, identity, relations, OR alternatives and substring containment. Exact values use year, count or value with a short string target. Do not generate answer options. Preserve strict/inclusive bounds, units, party roles and correct date/amount fields: payslip year means pay period; final total differs from balance due and individual line amounts. About a business does not mean issued by it.
 Choose each fact's scope from the requirement, not the extension or fact type:
 - any_unit: one matching page, slide, complete table, row, passage, frame or spoken segment establishes an occurrence. Nonmatching units do not disprove it. A flag on one PDF page, an animal somewhere in a video and words spoken somewhere are any_unit. Bind conditions that must concern the same page/row/object/passage/time in ONE question. Independent occurrences may be separate facts.
 - all_units: every relevant unit must pass, or a condition must be absent everywhere. A single counterexample disproves this; acceptance needs complete coverage. Identify the domain in the question, such as every page, every data row or every spoken passage. Silence is not a spoken passage; headers are not data rows. Do not add an existence condition unless requested.
@@ -19,7 +19,11 @@ The wording of the user’s container request does not belong in the local quest
 Every fact has a modality describing the evidence the verifier must inspect:
 - text: native machine-readable text, structured rows/cells, email fields, source code, or a speech transcript is authoritative. Use text for explicitly text-native formats such as TXT and Markdown, structured spreadsheets/CSV, email, code, and spoken-only requirements.
 - vision: pixels are authoritative. Use vision for visible objects, scenes, actions, layout, colors, diagrams, charts, handwriting, signatures, stamps, highlighting, and text visibly printed inside an image, video frame, rendered page, or rendered slide. An identity card or passport requested as a physical visual object, including a name visibly printed on it, is vision.
-- hybrid: native text may exist, but the requested document content can also be present only in a scan or embedded image. Use hybrid for textual or numeric facts in invoices, receipts, contracts, reports, tickets, certificates, Word files, PDFs, and unspecified general documents. This includes document identity, parties, dates, amounts, clauses, and exact phrases. Also use hybrid when a single fact genuinely needs both text or speech and visual evidence.
+- hybrid: native text may exist, but the requested document content can also be present only in a scan or embedded image. Use hybrid for textual or numeric facts in invoices, receipts, contracts, reports, tickets, certificates, Word files, PDFs, and unspecified general documents. This includes parties, dates, amounts, clauses, and exact phrases. Also use hybrid when a single fact genuinely needs both text or speech and visual evidence.
 Choose modality independently for every fact by asking what evidence can establish that fact. Search mode, query length, file extension, a person's name, and the presence of the word image/audio/video do not determine modality. Do not use hybrid as an uncertainty label.
 
 Treat a physical credential or other photographed physical document as a visual object. For example, checking whether an identity card bears a person's name is vision because the card and its printed fields must be seen. By contrast, an invoice addressed to a person is hybrid because it may be a native document or a scan. Spoken requirements in audio or video use text because the transcript is authoritative. A visible highlight, handwritten mark, page number, slide title position, chart, or diagram uses vision even when surrounding extracted text is long. Filename and file-metadata facts use text. A query may mix modalities across its facts.
+
+Use short natural questions, not "Does this evidence show...". Never add a standalone category check such as "Is this an invoice?", "Is this a contract?", "Are these meeting notes?", "Is this a PDF?", or their translations. Keep the requested content category in the useful question: "Is the final total of this invoice greater than 100 USD?" or "Does this contract allow automatic renewal?". Never discard a party, topic, subtype, role or other condition when removing a redundant category check. A pure lookup resolved by retrieval and metadata filters may have facts:[]; do not invent a question just to fill the list.
+Application sources: Apple Notes or Notes on a Mac maps to notes; Bear app to bear; Craft app to craft; Notion app to notion. Infer the source from the whole request, including app names mentioned at the end. An app name inside quoted content or the topic being discussed is not a source filter. Generic notes do not imply Apple Notes. App notes can contain typed text (text), images (vision), or unspecified content that may be either (hybrid).
+Preserve meaning across informal wording, typos and languages. A topic about a week does not imply a work week, creation date, or an unstated year. A short request may need complex verification; a long request may be a simple lookup. Do not add locations, dates, roles, amounts, or exclusivity that the request does not specify.