teamchong a52423870b dashboard: wire empirical cost fit into headline savings math + handoff
`DashboardState.fitCosts()` runs OLS over a rolling 50-event ring of
cold-miss (image_pixels, outgoing_text_chars, cache_create_tokens)
samples. `baselineCost()` and `estImageTokens()` consult the fit on
every event:

- `cost_fit != null` (N ≥ 3 samples) → headline saved_pct switches to
  the live empirical rates: txtReplaced = compressedChars × α,
  imgTokens = imagePixels × β.
- `cost_fit == null` → falls back to the stale constants
  (OPUS_IMAGE_TOKEN_COST = 2500, CHARS_PER_TOKEN = 4) — same behavior
  as before, so nothing breaks while data accumulates.

# Why a rolling ring, not seeded from history

Old events may carry the new instrumented fields but come from a
different renderer config (cell size, multiCol, atlas profile) or a
different model (Opus 4.6 vs 4.7 tokenizer). Mixing them in gives a
misleading α/β that lags reality. Each restart starts the fit empty
and refills from live traffic, so flipping `multiCol`, swapping
fonts, or upgrading the model auto-flushes stale samples.

# Endpoint

`/proxy-stats` JSON payload now carries:

  "cost_fit": {
    "alpha": 0.3, "beta": 0.0053,
    "chars_per_token": 3.3, "pixels_per_token": 189,
    "single_col_tokens_per_img": 4180,
    "multicol2_tokens_per_img": 8500,
    "n": 12
  }

null until ≥3 cold-miss samples land. Guards against degenerate fits
(collinear data, negative rates) by returning null rather than feeding
the dashboard noise.

# Math

Solves `tokens ≈ α · text_chars + β · pixels` via 2×2 normal
equations on the centered design matrix. Rounds α to 4 sig figs, keeps
β at 6 sig figs (β is tiny ~5e-3 so it needs the resolution).

# Open

Once N stabilizes at ~10+, re-grade the hardcoded
`OPUS_IMAGE_TOKEN_COST` and `CHARS_PER_TOKEN` constants to match the
empirical numbers. That just makes the fall-back path agree with
reality, doesn't change live behavior. HANDOFF.md captures the
verification recipe and priority queue for the next session.
2026-05-19 11:40:03 -04:00
2026-05-18 11:52:02 -04:00

pixelpipe

A token-saving proxy for Claude Code that renders the system prompt + tool definitions as images, so Claude OCRs them instead of paying for them as text. 65-73% input-token savings on Opus 4.7, 100% reasoning quality preserved, identical fixed text every turn for a clean prompt-cache.

Runs on Node 18+ and Cloudflare Workers from the same source.


How it works

                                  ┌─ original ────────────────────┐
                                  │ ~68K input tok                │
Claude Code  ──►  pixelpipe  ──►  │  (system + tools as text)     │  ──►  Anthropic
                       │          └───────────────────────────────┘
                       └──────►   ┌─ via proxy ───────────────────┐
                                  │ ~3.5K input tok               │
                                  │  (system + tools as PNG +     │
                                  │   prompt-cache breakpoint)    │
                                  └───────────────────────────────┘
                                          ↓ Anthropic vision OCR
                                          100% reasoning quality retained

The proxy intercepts POST /v1/messages, pulls the system prompt + tool documentation out of the JSON body, renders it into one or more grayscale PNGs using a build-time-generated GNU Unifont glyph atlas (covers ~35k BMP codepoints by default — Latin, Cyrillic, Greek, CJK, Hiragana, Katakana, Hangul, Hebrew, Arabic, math symbols, box drawing, decorative symbols), and substitutes those PNGs back in as image content blocks with an ephemeral cache_control breakpoint.

Token math (Opus 4.7, real Claude Code workflow):

metric original via proxy savings
Cold input tokens ~68K ~3.5K 95%
Cache-warm input tokens ~7.5K ~3.5K 53%
Per-call median (mixed) - - 65-73%
Per-image OCR quality vs txt - - ~99.5%

Quick start (Node)

npm install
npm run build           # produces dist/node.js
node bin/cli.js         # listens on 127.0.0.1:47821 by default

After editing code, restart in one step:

pnpm run restart                              # graceful SIGTERM of any running
                                               # instance → rebuild → fresh start
pnpm run restart -- --no-build                # skip rebuild (dist/ is fresh)
pnpm run restart -- --port 47822 --no-tools   # forward CLI flags to the proxy

pnpm run restart does, in order:

  1. Lists every running pixelpipe PID (via pgrep) and SIGTERMs them all. Orphans from prior crashed sessions are cleaned up too.
  2. Waits up to 5s for graceful exit (the SIGTERM handler flushes the JSONL tracker). Escalates to SIGKILL only if anything's still alive.
  3. Runs pnpm run build. Build failures abort the restart — the script refuses to start a stale binary. Pass --no-build to skip when you know dist/ is fresh.
  4. Checks the target port is free. If it isn't, names the holding process and refuses to start (cheaper than a crashed Node stacktrace).
  5. execs node bin/cli.js "$@" in the foreground so Ctrl-C reaches Node.

Point Claude Code at it:

ANTHROPIC_BASE_URL=http://127.0.0.1:47821 \
  claude --exclude-dynamic-system-prompt-sections

That's it. Use Claude Code normally.

The --exclude-dynamic-system-prompt-sections flag suppresses the small per-turn variable section so the rendered image stays byte-identical across turns — that's what makes the prompt cache actually hit.


Quick start (Cloudflare Workers)

npx wrangler dev        # local dev on :8787
npx wrangler deploy     # ship to *.workers.dev

Then in Claude Code:

ANTHROPIC_BASE_URL=https://pixelpipe.<your-account>.workers.dev \
  claude --exclude-dynamic-system-prompt-sections

You can attach a custom hostname and route in wrangler.toml.


Configuration

Both runtimes read the same options — Node from CLI flags or env, Worker from wrangler.toml [vars].

flag / var default meaning
--port PORT 47821 Node only — listen port
--upstream ANTHROPIC_UPSTREAM https://api.anthropic.com where to forward
--no-compress COMPRESS=0 on master switch
--no-tools COMPRESS_TOOLS=0 on fold tool docs into the image
--no-schemas COMPRESS_SCHEMAS=0 on include input_schema JSON in the image
--min-chars MIN_COMPRESS_CHARS 2000 skip compression below this many chars
--placement PLACEMENT system system or user — where image lands
--cols COLS 100 soft-wrap column count

In Workers, set the optional upstream API key with:

npx wrangler secret put ANTHROPIC_API_KEY

If unset, the proxy forwards whatever x-api-key the client sent.


Architecture

src/
├── core/              100% runtime-agnostic (Web Standard APIs only)
│   ├── atlas.ts         (generated) sparse Unicode atlas, base64-inlined
│   ├── png.ts           minimal grayscale PNG encoder
│   ├── render.ts        text → PNG bytes
│   ├── transform.ts     request body rewriter
│   ├── proxy.ts         the fetch handler
│   └── types.ts         Anthropic API types
├── node.ts            node:http adapter + CLI
└── worker.ts          export default { fetch }

scripts/
├── gen-atlas.ts       build-time: OTF → atlas.ts (uses @napi-rs/canvas)
└── build.mjs          esbuild bundler for Node target

assets/
├── Unifont-16.0.04.otf       primary font (~35k BMP codepoints w/ full-bmp profile)
├── UNIFONT_LICENSE.txt       OFL + GPL-with-font-exception
└── JetBrainsMono-Regular.ttf legacy / ASCII-only fallback (kept on disk)

The atlas is generated at build time from Unifont-16.0.04.otf, base64-inlined into a .ts file with sparse codepoint + offset tables (binary-packed), and shipped with the bundle. At runtime there are zero external files to read and zero non-Web-Standard imports — that's the only way this works in Workers without per-request asset fetches.

Regenerate the atlas (after swapping fonts, sizes, or codepoint profile):

pnpm run build:atlas                          # default: full-bmp (~35k cp, all BMP Unifont covers)
ATLAS_PROFILE=practical pnpm run build:atlas  # drops Hangul (~24k cp; for Workers free-tier)
FONT_PX=12 pnpm run build:atlas               # nondefault size; verify cells

Limitations

  • The bundled GNU Unifont at 10px (cell 5×11 px Latin, 10×11 CJK) is Anthropic-OCR-clean for ~35k BMP codepoints by default (full-bmp profile): Latin, Cyrillic, Greek, CJK Unified Ideographs, Hiragana, Katakana, Hangul, Hebrew, Arabic, math symbols, box-drawing, arrows, Dingbats, Letterlike Symbols, Enclosed Alphanumerics, etc. Drops for codepoints outside the profile (e.g. emoji 😀 — supplementary plane) get counted in events.jsonl#dropped_chars (with the top-20 broken out as dropped_codepoints_top) so you can spot patterns. For Workers free-tier deployments under the 1 MB compressed-bundle cap, switch to ATLAS_PROFILE=practical pnpm run build:atlas (~24k cp; drops Hangul). Right-to-left scripts render left-to-right in source order (no bidi shaping); Devanagari / Thai / similar complex shaping is also unsupported.
  • Compression sets a 5-minute prompt-cache TTL. Adding cache_control: ephemeral causes warm-cache rotation, not eviction.
  • A 5KB break-even point: if input is < MIN_COMPRESS_CHARS chars we skip compression entirely (overhead would exceed savings).
  • Per-machine font: regenerate the atlas if you swap fonts. The generated src/core/atlas.ts is checked in so consumers don't need @napi-rs/canvas to install.
  • Workers CPU limit: this is fine for free-tier (10ms CPU) on small prompts; large prompts (>30K chars) may need the paid tier.

Development

npm install
npm run dev:node              # tsx watch on src/node.ts
npm run dev:worker            # wrangler dev
npm run test                  # vitest
npm run test:watch
npm run typecheck             # tsc --noEmit
pnpm run build:atlas          # regenerate src/core/atlas.ts from OTF
npm run build                 # build dist/node.js
npm run deploy:worker         # wrangler deploy

License

MIT.

S
Description
cut Fable 5 token usage by rendering text context as images
Readme MIT 24 MiB
Languages
TypeScript 86.7%
JavaScript 9.5%
Shell 2.1%
Python 1.7%