Colour Memory

Generate a complete multi-room house colour scheme from a single archive. Where the retrieved evidence supports it, rooms are grouped by documented provenance (same object, same pattern colourway, same source citation, same designer, same period+institution); where it does not, they are grouped by perceptual similarity and the response says so -- grouping_basis reports which, and provenance_credit reports how many delivered colours are individually evidenced versus documented as a GROUP (often zero). Every colour is a real archive record either way. Each room gets a full colour card set (dominant/secondary/joinery/textile/accent roles) with paint matches, LRV, illuminant behaviour, and WCAG data. A documentary grouping states exactly which documented fact ties its colours together (e.g. 'documented on the same Fitzwilliam Museum record for the Artichoke hanging, accession T.1-1979') and its weighted grouping_strength score -- historical claims are templated only from fields the archive record actually states; room/surface assignment is disclosed as editorial, never as historical fact. Orientation (north/south/east/west) is recorded as room context and surfaced in each room's light_context; in the current model it does not alter palette selection. Example: concept='Arts and Crafts movement family house', archive='ArtsAndCrafts', rooms=[{label:'Entrance Hall', room_type:'hallway', orientation:'north'}, {label:'Living Room', room_type:'living', orientation:'south'}]. Use interior_specify instead for a single room without cross-room grouping. The response includes house_ref -- pass that (not the full response) to palette_pdf's house_ref input to render the flagship document; the full response is often 60-100KB+, too large to safely resend as another tool call's input. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you a…

Compound endpoint: one image in, full creative brief out. Extracts dominant colours, matches them to the Colour Memory archive with coherent cultural naming, understands the scene, identifies the style period, suggests product directions by category (textiles, interiors, fashion), generates an image generation prompt, and returns a swatch URL. Use instead of chaining palette_extract + palette_analyse + agent_brief separately. Pass style_context for coherent archive matching e.g. 'English cottage garden', 'Victorian', 'MarsColour'. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Upload an image (base64 encoded) and extract its dominant colour palette, with each colour matched to its nearest named archive entry with full cultural provenance. Uses K-means++ extraction; sRGB pixels are read as D65 CIELAB for matching. Returns up to 5 dominant colours, each with archive name, cultural story, nearest RAL standard, and WCAG accessibility data. Works for product photography, interior photos, artwork, brand assets, and mood boards. The image is never stored — processed in memory only. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Upload a portrait photo and receive a full personal colour analysis. Determines your seasonal type (Spring, Summer, Autumn, or Winter), colour depth (light, medium, or deep), and undertone (warm, cool, or neutral). Returns a curated palette of archive colours that genuinely suit you — each with full historical provenance and cultural context — plus colours to avoid. Uses Claude Vision for skin, hair, and eye analysis, then matches to the archive by CIEDE2000 perceptual distance. The photo is never stored. Example: a Deep Winter might wear Ottoman Carbon Ink while a True Spring suits Kogi Mango.

Colour Memory's proprietary semantic metric. Score how tightly the material origin of a colour aligns with its social consequence. 1.00 = material and consequence are indistinguishable (blood as prognosis, ash as finality). 0.80 = institution mediates the colour (paint as deterrence, flag as authority). 0.50 = symbolic or associative only. Input: list of colour entries with name, hex, archive, source, notes. Output: resonance score, material origin, social function, alignment reason, confidence. Use for investigative reports, forensic briefs, museum content, editorial PDFs. This is the metric that separates Colour Memory from palette generators.

Upload an image to Colour Memory for ephemeral processing. Returns image_id valid 5 minutes. Pass image_id to palette_extract or image_brief. Use for any image over 50 KB.

Generate a complete interior colour specification from a concept or brief. Input a room concept, type, and style — receive a professionally structured colour scheme with 60/30/10 surface assignments, archive colour names with full cultural provenance, Farrow and Ball and Little Greene paint matches, three-illuminant light behaviour (D65 daylight, F11 atrium, Illuminant A incandescent), WCAG accessibility for digital use, and a written cultural rationale explaining why each colour belongs in this room. Examples: 'bold maximalist living room', 'calm Scandi bedroom', 'Victorian study', 'coastal kitchen', 'gallery hallway'. Pass `archive` to lock the whole scheme to one named archive (e.g. 'ArtsAndCrafts', 'GeorgianPleasures') -- every returned colour will belong to it, or you get a clear error listing valid archive names rather than a silent unrestricted search. Use /interior-specification/pdf for a downloadable branded PDF version. Use house_specify instead for a multi-room scheme with cross-room provenance grouping. This is the tool that replaces a colour consultation. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Return a live inventory of all active endpoints and MCP tools. Use this first to discover what the API can do before making calls. Returns tool count, endpoint list, MCP-exposed tools, and usage notes. Deterministic -- no LLM cost.

Analyse a palette of hex colours against the Colour Memory archive. For each colour returns the nearest named archive entry with cultural name, source, claim_strength (A-E), do_not_say guardrails, and alternatives when confidence is low. Palette-level deduplication ensures no two colours map to the same archive entry. Use after extracting colours from a photo or generating a palette. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Full palette quality audit. Scores on accessibility, cultural risk, tonal balance, colour diversity, and archive naming strength. Returns overall score 0-100, grade, and prioritised fix list. Enterprise quality gate -- use before shipping any palette. Deterministic, no LLM cost. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Deep perceptual, cultural, and commercial comparison between two palettes. Returns timelessness scores, commercial strength, cultural depth, emotional difference, and a winner verdict for the stated use case. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Generate a historically grounded colour palette from a cultural concept or theme. Returns 4-6 coordinated archive colours with hex values, proportions, and provenance. Examples: 'Victorian mourning', 'Ottoman court', 'Japanese wabi-sabi', 'Scandinavian winter', 'West African kente', 'Renaissance Florence'. Every colour returned is sourced from the archive with documented history. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Export a palette to CSS custom properties, Figma design tokens, Tailwind config, ASE hex list, or JSON. Each colour is automatically named from the archive. Embeds Colour Memory directly into design workflows.

Extract dominant colours from an image using k-means++ clustering. Accepts a public image URL or base64-encoded image. Returns hex values with proportional weights sorted by luminance. Optionally runs palette_analyse on the results. Use this instead of image_palette when you need hex values with proportions for palette_analyse or palette_swatch. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Package a palette YOU HAVE ALREADY CHOSEN into a professional, visible, downloadable deliverable. Call this ONCE, after you have decided the final colours for the customer -- it verifies and packages your decision and returns the palette image plus PNG, SVG, PDF, ASE, JSON and CSS. It does not search, select, replace or improve anything: the colours you send are the colours that come back, in your order. Send every final colour with its name, and mark a colour source-backed only if it really is a Colour Memory archive record -- anything unverified is presented as your recommendation, which is the honest description of it.

Send a palette of up to 12 slots, locking some with hex values and leaving others empty. Empty slots are filled with the nearest CIEDE2000 archive match, interpolated from the locked anchors. Optional archive filter restricts fills to one archive. Returns full citation — name, archive, primary source, colour notes — for every filled slot. Example: lock a client's existing wall colour and fill a 5-colour scheme from Oxfordshire. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Generate a perceptually smooth gradient between 2-5 archive anchor colours. Each interpolated stop snaps to the nearest real archive colour by CIEDE2000. Anchor stops are kept true to their source. Choose linear (physically accurate Lab interpolation) or chroma_preserved (LCh interpolation, short-arc hue, avoids desaturated midpoints). Returns stop array, CSS linear-gradient string, or SVG swatch bar. Use for design briefs, colour journey visualisations, and gradient systems. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Given a legacy palette, generate an archive-grounded premium support system. For each existing colour: identifies its historical archive anchor, names it, and scores its provenance confidence. Detects palette gaps and fills them from the archive. Returns full palette with roles, confidence scores, CSS tokens, and production notes. Every addition has a named historical origin. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Refine an existing palette using natural language feedback. Submit your current palette and feedback such as more melancholic, too corporate add warmth, or better for Gen Z luxury. Returns a refined palette with archive grounding and change rationale. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Generate light-mode and dark-mode role maps from a palette. Analyses LRV, assigns background/surface/text/accent roles for each mode, checks body text contrast safety, and flags missing neutrals. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Generate a premium branded PDF specification sheet from a palette of archive entries. Returns a downloadable PDF with full-bleed colour panels, archive names, provenance notes, RAL nearest match, LRV, chroma, WCAG contrast data, and Colour Memory branding. Use this to create client deliverables, specification sheets, and print assets. Two supported entry sources: (1) the flat entries array from query_hex or palette_from_concept -- each item needs name, hex, archive_source, colour_notes, primary_source, zone; or (2) archive_report_brief's output -- pass report['presented_cards'] as entries, NOT report['colour_cards'] (the legacy field is kept only for backward compatibility and does not carry family-card presentation). presented_cards may mix three card shapes, all rendered as exactly one page each regardless of how many underlying records they represent: single_cultivar (a normal flat card), colour_family (card_type='colour_family', many near-identical records summarised as one card with member_count/supported_by/family_delta_e_max), and conservative-mode hero cards (a single_cultivar card carrying family_grouping_suppressed=true and also_supported_by, representing a larger natural family that was folded into one promoted cultivar instead of shown as a group). delivery controls how the PDF is returned: 'url' (default) gives a signed, time-limited R2 download link; 'base64' returns the PDF bytes inline in the response for callers that need the file directly rather than a link. Alternative input mode: pass `sections` instead of `entries` for a grouped, multi-section document (e.g. house_specify's rooms) instead of one page per flat colour card. Each section has a title, rationale, and its own entries. `presentation.template` picks the layout: 'sectioned_specification' (clean grid-per-section spec sheet) or 'colour_drenched_lookbook' (hero colour fills the page, swatches cover most of it, citations collected into a final appendix rather than shrunk onto every page). T…

Generate a complete interior specification from 2-8 hex values. Returns surface assignments, 60-30-10 proportions, lighting behaviour, and archive colour names. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Like palette_concept but with archive filtering and relevance controls. Use allowed_archives to restrict results to specific cultural traditions e.g. ['Japan'] for Japanese only. Use min_relevance to filter weak concept matches. Fixes cross-archive drift when cultural specificity matters. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Generate a clean, text-free PNG swatch image from hex colours. Returns a URL to the PNG. Use for Midjourney --sref style references or design mood boards. Supports photo-proportional weights from palette extraction, equal distribution, grid layout, a true smooth LCh-interpolated gradient (no hard colour edges, best for mood/atmosphere/colour-grade references rather than literal composition), and 13 fixed design ratios (6310, 7020, triptych, quad, filmstrip, etc.).

Map any list of hex values into a target archive using CIEDE2000 nearest-neighbour matching. Each input hex is matched to the closest named colour in the chosen archive, with a delta-e relevance band (exact / close / approximate / loose) and full provenance. Use to translate a client's paint colours into Shakespeare language, map a brand palette into historical Japanese pigments, or find the nearest Oxfordshire equivalents to a French scheme. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Evaluate a palette of 2-8 hex values for a use case, market, and medium. Returns a verdict (strong / strong_with_adjustment / weak / avoid), a score 0-100, the role of each colour, the single biggest weakness, and a concrete suggestion for what to add to fix it. Each colour is matched to the nearest archive entry for cultural grounding. Examples: 'premium cushion collection UK ecommerce', 'hotel lobby interior', 'SaaS brand identity global digital'. The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Soft-delete a durable house project you own. Idempotent -- deleting an already-deleted, unknown, or not-owned project_id returns the same successful-shaped response either way, and never discloses which case occurred. Version rows are never physically erased. Once deleted, a project can no longer be resolved by project_id or house_ref.

Export the same house-flagship content palette_pdf renders as a genuinely editable Word document or Excel workbook -- for a client who needs to annotate the scheme, or a contractor building a materials/procurement schedule from it. Accepts exactly one of house, house_ref, or project_id (same resolution and ownership rules as palette_pdf: an authenticated project_id or house_ref requires its owner; an anonymous temporary house_ref works bearer-only). Never calls house_specify, never re-runs reasoning or re-matches archives -- it renders exactly what was already resolved and stored. DOCX mirrors the PDF flagship's structure (title page, house thesis, whole-house progression, one section per room, implementation details with the same historical/digital-estimate/commercial-candidate distinction, and a deduplicated source appendix) as real editable paragraphs and tables, never images. XLSX produces three sheets: Finishes Schedule (one row per room-surface-colour assignment), Evidence Register (one row per unique colour, the full citation/claim-strength/do_not_say trail), and Paint Procurement (one row per unique paint product, deduplicated for ordering). Every field left blank where the underlying stored data doesn't have it -- nothing is invented. delivery="url" (default) gives a signed R2 download link; delivery="base64" returns the file inline unless the encoded result is too large, in which case it fails closed with OUTPUT_TOO_LARGE_FOR_INLINE_DELIVERY and a repair instruction to use delivery="url" instead.

Retrieve summary metadata for a durable house project you own -- title, project type, current version number, schema/archive revision, created/updated timestamps, and the content hash of the requested version. Requires an API key; the project must belong to the calling key. Does NOT return the full house specification (rooms, colours, sources) -- use house_ref with palette_pdf to render a document from a project, or project_versions to see its version history. An unknown, deleted, expired, or not-owned project_id all return the same generic not-found response, by design.

List durable house projects owned by the calling API key, most recently updated first. Anonymous (keyless) calls always return an empty list -- temporary anonymous projects are never listable, only resolvable directly by their house_ref. Requires an API key.

List every stored version of a durable house project you own, oldest first, with each version's content hash, build/schema/archive revision, and creation time. Version 1 is always immutable and always present. Earlier versions remain renderable via palette_pdf's project_version input even after a newer version exists. Requires an API key and ownership.

Ask a cultural, historical, or material colour question. Returns named archive colours with provenance and cultural context. Works for abstract queries like 'grief' (returns Ophelia Violet, Titan's Grief), 'toxic Victorian pigments' (returns Nightshade Poison Berry Purple, Victorian Arsenic Mirror Silver), or 'the sea at dusk' (returns Evening Sea, Thames Blue Dusk). Even a vague or personal-sounding query never comes back empty -- result_path in the response tells you honestly whether this was a 'direct' hit, a 'broadened' retry (relaxed match threshold), or a 'redirected' fallback (dropped an archive restriction or switched to full-corpus semantic search).

Find the closest named archive colours to a hex value using CIEDE2000 perceptual distance.

Show a finished colour result to the customer AS COLOUR. Call this last, once the palette, house scheme, brand system or audit you are going to hand over has been decided -- it presents what you already have and chooses nothing. Do not call it for a single contrast figure, a colour history question with no palette, or tool guidance. Send the story with it: one sentence on what the scheme IS, the design decision you made, and the one rule worth remembering. The colours cannot carry those on their own.

Single-call forensic colour brief generator. Runs coverage gap analysis, pulls best archive colours, checks for anachronisms, scores claim roles (anchor/support/analogue/provocation/reject), auto-rejects stubs, generates editorial argument, act structure, pull quote, closing line, and image prompt via Claude. This replaces chaining coverage_gap + archive_report_brief + anachronism_guard + resonance_index + evidence_gap separately. Input: title, audience, themes, archives, period, tone. Output: complete deliverable package ready for PDF or editorial use. Tone options: forensic (default), editorial, clinical, narrative.

The colour question every stylist gets asked: does this bag go with this outfit? Submit your outfit items as hex values with labels (dress, bag, shoes, coat, belt, scarf, etc.) and receive a verdict on what works, what clashes, what is missing, and what to add. Every recommendation is backed by archive colour names and historical context — not generic colour theory, but documented cultural combinations. Also suggests one missing archive colour that would complete the look. Examples: 'I have a navy dress (#1C3A6E) and a tan bag (#C8A87A) — what shoes?' or 'Does this burgundy coat work with olive trousers?' The result already carries the rendered palette and its PNG, PDF, ASE, JSON and CSS downloads -- show them to the customer. Never present the archive anchors a colour was derived from as the colours you are recommending. If you go on to choose a final palette OF YOUR OWN from this evidence, call palette_finalize once with those exact colours so the customer can see and download what you actually recommended.

Zero-argument diagnostic call: returns exactly what build is live right now (build_sha, schema_version, schema_hash, archive_revision, tool_count, generated_at, server_environment). Use this to confirm you're seeing the current deployed schema -- e.g. if a tool you expect (like house_specify) seems to be missing, call this first and compare schema_hash against another connection before assuming the server is wrong; a mismatch usually means one connection is holding a stale cached manifest, not a real deployment gap.

Deterministic tool-routing lookup -- not an LLM call, always instant, never consumes archive or LLM credit. Give it a goal in plain words (e.g. 'I have a brand colour and want to check it against competitors', 'I want to paint a whole house', 'I have a photo and want the colours from it') and it returns exactly which tool to call first, what to call next, and which superficially-similar tools NOT to start with and why. Use this before guessing between similarly-named tools (e.g. palette_specify vs interior_specify vs house_specify, or colour_passport vs colour_dna).

Generate a complete WCAG-compliant UI state palette from a brand hex. Returns colours for: brand, hover, active, disabled, focus ring, success, warning, error, info, surface subtle, surface strong. All states computed for contrast against your background colour. Returns hex, contrast ratio, WCAG grade, and usage note for each state. Includes CSS custom properties ready to paste. Supports light and dark mode. Use before building any UI component system.

Emily Noyes Vanderpoel's documented 'good' dyad/triad hue groupings (Color Problems, 1902, Plate XXXVI 'Good Dyads or Pairs' and Plate XXXVII 'Good Triads'), served from the build artefacts vanderpoel_dyads.json and vanderpoel_triads.json. Snaps an input hex to the nearest of Vanderpoel's eleven real hue-family labels, resolves that hex against the archive, and returns every documented dyad or triad the label belongs to -- each OTHER member carrying its own nearest real archive match (CIEDE2000), with hex_collision disclosed when the match is tied. Unlike hay_contrast, this carries NO proportional ratio -- only Vanderpoel's own categorical 'good' editorial judgement, printed verbatim on the plates. Returns groupings: null for a near-neutral input, where hue is not meaningfully defined, or when the nearest hue-family label is a member of no dyad or triad -- silence is the honest answer, not a forced match. Member names come from a transcriber's digitisation_alt_text (an accessibility description of the plate image), not verified printed captions; every response states this caveat explicitly (member_source_caveat), not just a code comment. The hue positions used to snap an input to a label are a modern matching convention, not a number Vanderpoel printed -- disclosed per response as label_placement_basis.

Are these two specific archive records connected -- and if so, exactly why? Traces the real path between them through Colour Memory's stored relationship graph (e.g. two Wada colours that share membership in the same published combination) and returns the actual path, not just a yes/no. States plainly when two records are NOT connected, and always states what a shared connection does not prove -- e.g. that two colours sharing a Wada combination were independently asserted as a pair, not just co-published in the same numbered plate. Get record_id values from archive_search or colour_card.