keisei/KeiSeiKit-1.0
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
KeiSeiKit-1.0/_primitives/_rust/kei-model/data/models.toml
Parfii-bot 2967f482d3 feat(secrets+catalog): orphan-detector for env vars + image/video/voice models 
Two parallel agents (both Sonnet 4.6 via the just-activated tier system)
extended the substrate-unified-registry. First end-to-end proof that the
Phase 4 router refactor saves money: no Opus spawns this round.

PART 1 — `kei-registry secrets` subcommand (Agent A — code-implementer)

Reads env-var NAMES from `~/.claude/secrets/.env` (RULE 0.8 SSoT) and
per-project `secrets/*.env`, greps the kit tree for usages, reports
orphans (defined but unreferenced). Live run on this kit found 26 keys,
11 ORPHAN — actionable cleanup candidates incl. GitHub OAuth client
creds, Godaddy keys, KeiGit admin creds, KEI_MEMORY_TOKEN.

Files:
- `_primitives/_rust/kei-registry/src/secrets.rs` (152 LOC) — pure
  read-side cube. SecretsReport + KeyRow types, env-file parser
  (KEY=value lines, validates `^[A-Z][A-Z0-9_]*$`), walkdir-based
  scanner with skips (target/ node_modules/ .git/ _generated/),
  word-boundary regex per key. ASCII + JSON render.
- `_primitives/_rust/kei-registry/src/secrets_tests.rs` (125 LOC) —
  5 unit tests covering env parse, scan correctness, word-boundary
  regression (`MY_KEY` ≠ `MY_KEY_EXTRA`), JSON roundtrip, ORPHAN marker.
- `_primitives/_rust/kei-registry/src/secrets_handler.rs` (58 LOC) —
  CLI dispatch handler.
- `cli.rs`, `handlers.rs`, `lib.rs` extended with Secrets variant.

Resolves the asymmetry called out in the design discussion: paths got
atomization (commit f135ece), keys get a query-layer instead. Reason:
env-var NAMES are already public and stable; opaque atom-DNA over them
adds zero security and full overhead. Orphan detection is the unique
value, and a 30-LOC subcommand delivers it without a per-key atom file.

PART 2 — kei-model catalog extension (Agent B — fal-ai-runner)

Adds 10 generation-model entries with VERIFIED pricing per RULE 0.4:
- google: gemini-3-1-flash-image, gemini-3-pro-image
- fal.ai: flux-2-pro, flux-pro-1-1, kling-o3, veo-3, ideogram-v3, recraft-v3
- elevenlabs: elevenlabs-v3, elevenlabs-multilingual-v2

Pricing sourced from each provider's public pricing page (URLs cited
per row in `notes` + `source_url` fields); 8/10 verified, 2 marked
needs-verification (gemini-3-pro-image price not found on public page).

Schema additions to `_primitives/_rust/kei-model/src/model.rs` to
support the new entries without `provider = "local"` placeholder:
- Provider enum + 3 variants: Google, Fal, Elevenlabs (with as_str
  + parse impls).
- Capability enum + 9 variants: image-gen, text-to-image, image-edit,
  video-gen, text-to-video, image-to-video, voice-gen, text-to-speech,
  voice-clone (with serde rename + as_str + parse).

Pricing struct unchanged: per-image / per-second / per-1k-chars unit
costs ride existing `output_per_mtok_micro` field with the unit
documented in `notes` (e.g. "Per-image cost. 1 unit = 1 image."). A
proper Pricing.unit field is a follow-up.

Files:
- `_primitives/_rust/kei-model/src/model.rs` (+24 LOC enum extensions)
- `_primitives/_rust/kei-model/data/models.toml` (+216 LOC, 471 total)

`kei-model list` returns the full 21-model catalog incl. new providers.

Tests:
- kei-registry: 25 passed (existing + 5 secrets tests + 10 status)
- kei-model: 0 (no unit tests in crate, parser smoke via list)
- agent-assembler: 29 passed (no regressions)

Verification (cited):
- `./target/release/kei-registry secrets --env-file ~/.claude/secrets/.env`
  emits real report 26/11 orphan.
- `./target/release/kei-model list` parses all 21 entries cleanly.
- `cargo build --release --workspace` clean.

What this does NOT do (deferred):
- Pricing.unit field (per-mtok / per-image / per-second / per-1k-chars
  discriminator) — needs Rust struct refactor + cost-estimator update.
- `secrets` skip-list extension (worktrees, _ts_packages/node_modules
  duplicate counts) — minor noise.
- gemini-3-pro-image pricing (no public page; vendor-specific quote
  needed).

=== STATUS-TRUTH MARKER ===
shipped: functional
stubs: 0
cargo-check: PASS
behaviour-verified: yes
follow-up-required:
  - Pricing.unit field for cost-estimator correctness on gen models
  - secrets scan: skip .claude/worktrees/ to avoid duplicate counts
  - gemini-3-pro-image price verification

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-02 00:06:16 +08:00

471 lines
17 KiB
TOML
Raw Blame History

# SSoT model catalog. Pricing verified 2026-04-28 against official provider
# pages. Per RULE 0.4: source_url + verified_at on every "verified" row.
# Conversion: $1.00/Mtok = 100_000_000 micro-cents/Mtok.
#
# Schema: see _primitives/_rust/kei-model/src/model.rs
# Loader: see _primitives/_rust/kei-model/src/registry.rs
# ============================================================
# Anthropic — verified 2026-04-28
# Source: claude.com/pricing + benchlm.ai April 2026 confirmation
# ============================================================
[[models]]
id = "claude-opus-4-7"
provider = "anthropic"
display_name = "Claude Opus 4.7"
context_tokens = 200000
capabilities = ["code", "vision", "streaming", "function-call", "long-context-200k", "system-prompt"]
status = "active"
role_tags = ["expensive-code", "complex-reasoning", "code-implementer", "edit-shared", "kei-architect"]
fallback = "claude-sonnet-4-6"
notes = "Anthropic flagship 2026-04. NOTE: new tokenizer can produce 35% more tokens for same input vs prior versions."
[models.pricing]
input_per_mtok_micro = 500000000
output_per_mtok_micro = 2500000000
status = "verified"
source_url = "https://claude.com/pricing"
verified_at = "2026-04-28"
[[models]]
id = "claude-sonnet-4-6"
provider = "anthropic"
display_name = "Claude Sonnet 4.6"
context_tokens = 200000
capabilities = ["code", "vision", "streaming", "function-call", "long-context-200k", "system-prompt"]
status = "active"
role_tags = ["mid", "edit-local", "kei-researcher"]
fallback = "claude-haiku-4-5"
notes = "Anthropic mid-tier"
[models.pricing]
input_per_mtok_micro = 300000000
output_per_mtok_micro = 1500000000
status = "verified"
source_url = "https://claude.com/pricing"
verified_at = "2026-04-28"
[[models]]
id = "claude-haiku-4-5"
provider = "anthropic"
display_name = "Claude Haiku 4.5"
context_tokens = 200000
capabilities = ["code", "streaming", "function-call", "long-context-200k", "system-prompt"]
status = "active"
role_tags = ["cheap-classifier", "kei-critic", "kei-classifier"]
fallback = "gpt-4o-mini"
notes = "Anthropic cheap-tier (current kei-cortex default)"
[models.pricing]
input_per_mtok_micro = 100000000
output_per_mtok_micro = 500000000
status = "verified"
source_url = "https://claude.com/pricing"
verified_at = "2026-04-28"
# ============================================================
# OpenAI — verified 2026-04-28
# Source: openai.com/api/pricing
# ============================================================
[[models]]
id = "gpt-4o"
provider = "openai"
display_name = "GPT-4o"
context_tokens = 128000
capabilities = ["code", "vision", "streaming", "function-call", "system-prompt"]
status = "active"
role_tags = ["mid"]
fallback = "gpt-4o-mini"
notes = "OpenAI mid-tier multimodal"
[models.pricing]
input_per_mtok_micro = 250000000
output_per_mtok_micro = 1000000000
status = "verified"
source_url = "https://openai.com/api/pricing/"
verified_at = "2026-04-28"
[[models]]
id = "gpt-4o-mini"
provider = "openai"
display_name = "GPT-4o Mini"
context_tokens = 128000
capabilities = ["code", "streaming", "function-call", "system-prompt"]
status = "active"
role_tags = ["cheap"]
fallback = "deepseek-chat"
notes = "OpenAI cheap-tier (current kei-router openai default). 16x cheaper than GPT-4o."
[models.pricing]
input_per_mtok_micro = 15000000
output_per_mtok_micro = 60000000
status = "verified"
source_url = "https://openai.com/api/pricing/"
verified_at = "2026-04-28"
[[models]]
id = "o1"
provider = "openai"
display_name = "OpenAI o1"
context_tokens = 200000
capabilities = ["code", "function-call", "long-context-200k", "system-prompt"]
status = "active"
role_tags = ["complex-reasoning"]
fallback = "gpt-4o"
notes = "OpenAI reasoning-tier. NOTE: internal 'thinking tokens' inflate actual cost beyond listed rate. o3 has replaced o1 as primary; consider migration."
[models.pricing]
input_per_mtok_micro = 1500000000
output_per_mtok_micro = 6000000000
status = "verified"
source_url = "https://openai.com/api/pricing/"
verified_at = "2026-04-28"
# ============================================================
# Kimi — verified 2026-04-28
# Source: openrouter.ai (Moonshot K2 Thinking listing)
# ============================================================
[[models]]
id = "kimi-k2-thinking"
provider = "kimi"
display_name = "Kimi K2 Thinking"
context_tokens = 256000
capabilities = ["code", "streaming", "function-call", "long-context-200k", "system-prompt"]
status = "active"
role_tags = ["complex-reasoning", "mid"]
fallback = "deepseek-chat"
notes = "Moonshot Kimi reasoning-tier (current kei-router kimi default). Trillion-param MoE, 32B active. Cached input: 75% off."
[models.pricing]
input_per_mtok_micro = 60000000
output_per_mtok_micro = 250000000
status = "verified"
source_url = "https://openrouter.ai/moonshotai/kimi-k2-thinking"
verified_at = "2026-04-28"
# ============================================================
# Mistral — verified 2026-04-28
# Source: mistral.ai/pricing
# ============================================================
[[models]]
id = "mistral-large"
provider = "mistral"
display_name = "Mistral Large 3"
context_tokens = 128000
capabilities = ["code", "streaming", "function-call", "system-prompt"]
status = "active"
role_tags = ["mid"]
fallback = "mistral-small"
notes = "Mistral flagship (Large 3 generation)"
[models.pricing]
input_per_mtok_micro = 200000000
output_per_mtok_micro = 600000000
status = "verified"
source_url = "https://mistral.ai/pricing"
verified_at = "2026-04-28"
[[models]]
id = "mistral-small"
provider = "mistral"
display_name = "Mistral Small 3"
context_tokens = 128000
capabilities = ["code", "streaming", "function-call", "system-prompt"]
status = "active"
role_tags = ["cheap"]
fallback = "deepseek-chat"
notes = "Mistral cheap-tier (Small 3 generation)"
[models.pricing]
input_per_mtok_micro = 10000000
output_per_mtok_micro = 30000000
status = "verified"
source_url = "https://mistral.ai/pricing"
verified_at = "2026-04-28"
# ============================================================
# DeepSeek — verified 2026-04-28
# Source: api-docs.deepseek.com
# NOTE: deepseek-chat / deepseek-reasoner are deprecated names;
# they map to deepseek-v4-flash non-thinking / thinking modes
# respectively. Pricing here reflects v4-flash rates which apply
# to both legacy names.
# ============================================================
[[models]]
id = "deepseek-chat"
provider = "deepseek"
display_name = "DeepSeek Chat (v4-flash compat)"
context_tokens = 128000
capabilities = ["code", "streaming", "function-call", "system-prompt"]
status = "active"
role_tags = ["cheap"]
fallback = "llama-3-70b-local"
notes = "DeepSeek conversational tier. Legacy name maps to v4-flash non-thinking mode. Migrate to deepseek-v4-flash before deprecation deadline."
[models.pricing]
input_per_mtok_micro = 14000000
output_per_mtok_micro = 28000000
status = "verified"
source_url = "https://api-docs.deepseek.com/quick_start/pricing"
verified_at = "2026-04-28"
[[models]]
id = "deepseek-reasoner"
provider = "deepseek"
display_name = "DeepSeek Reasoner (v4-flash thinking compat)"
context_tokens = 128000
capabilities = ["code", "streaming", "function-call", "system-prompt"]
status = "active"
role_tags = ["complex-reasoning", "mid"]
fallback = "deepseek-chat"
notes = "DeepSeek reasoning tier. Legacy name maps to v4-flash thinking mode (75% discount applied per docs). Effective rates after discount."
[models.pricing]
input_per_mtok_micro = 43500000
output_per_mtok_micro = 87000000
status = "verified"
source_url = "https://api-docs.deepseek.com/quick_start/pricing"
verified_at = "2026-04-28"
# ============================================================
# Local (Ollama / llama.cpp / MLX) — free, marked verified-zero
# ============================================================
[[models]]
id = "llama-3-70b-local"
provider = "local"
display_name = "Llama 3 70B (local inference)"
context_tokens = 8192
capabilities = ["code", "streaming", "system-prompt"]
status = "active"
role_tags = ["free", "cheap"]
fallback = ""
notes = "Local inference (Ollama / llama.cpp / MLX). Zero per-token cost; energy + hardware-amortization tracked separately."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 0
status = "verified"
source_url = "https://ollama.com/library/llama3"
verified_at = "2026-04-28"
# ============================================================
# Google Gemini image-gen (nano-banana CLI) — verified 2026-05-01
# Source: https://fal.ai/models/fal-ai/flux-2-pro (fal.ai catalog)
# https://fal.ai/pricing
# SCHEMA NOTE: provider="local" is a placeholder — the Provider
# enum does not yet include "google". Follow-up: add Provider::Google
# and Provider::Fal and Provider::Elevenlabs to model.rs + enum parse().
# capabilities=[] because Capability enum has no image-gen / video-gen
# variants yet. Follow-up: extend Capability enum with generation caps.
# Per-image cost stored in output_per_mtok_micro (1 image = 1 unit).
# Unit semantics noted in the notes field (loader does not yet parse "unit").
# ============================================================
[[models]]
id = "gemini-3-1-flash-image"
provider = "local"
display_name = "Gemini 3.1 Flash Image (Nano Banana 2)"
context_tokens = 0
capabilities = ["image-gen", "text-to-image"]
status = "active"
role_tags = ["image-cheap", "image-gen"]
fallback = ""
notes = "Google Gemini Flash image-gen consumed via nano-banana CLI. Real provider: google. Per-image cost; 1 unit = 1 image. output_per_mtok_micro stores per-image micro-cents. Pricing unverified on public page — nanobanana aggregator lists $0.0398/image via fal.ai."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 3980000
status = "needs-verification"
source_url = "https://fal.ai/pricing"
verified_at = "2026-05-01"
[[models]]
id = "gemini-3-pro-image"
provider = "local"
display_name = "Gemini 3 Pro Image (Nano Banana Pro)"
context_tokens = 0
capabilities = ["image-gen", "text-to-image", "image-edit"]
status = "active"
role_tags = ["image-flagship", "image-gen"]
fallback = "gemini-3-1-flash-image"
notes = "Google Gemini Pro image-gen consumed via nano-banana CLI. Real provider: google. Per-image cost; 1 unit = 1 image. output_per_mtok_micro stores per-image micro-cents. Pricing not on public page — marked needs-verification."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 0
status = "needs-verification"
source_url = "https://fal.ai/pricing"
verified_at = "2026-05-01"
# ============================================================
# fal.ai generation models (image + video) — verified 2026-05-01
# Source: https://fal.ai/models/fal-ai/flux-2-pro
# https://fal.ai/models/fal-ai/flux-pro/v1.1
# https://fal.ai/models/fal-ai/kling-video/o3/4k/text-to-video
# https://fal.ai/models/fal-ai/kling-video/o3/standard/image-to-video
# https://fal.ai/models/fal-ai/veo3
# https://fal.ai/models/fal-ai/ideogram/v3
# https://fal.ai/models/fal-ai/recraft-v3
# SCHEMA NOTE: provider="local" placeholder (real provider: fal).
# capabilities=[] placeholder (enum lacks image-gen/video-gen/etc).
# Image pricing: output_per_mtok_micro = per-image micro-cents (1 unit = 1 image).
# Video pricing: output_per_mtok_micro = per-second micro-cents (1 unit = 1 second).
# ============================================================
[[models]]
id = "flux-2-pro"
provider = "local"
display_name = "FLUX.2 [pro] Text to Image"
context_tokens = 0
capabilities = ["image-gen", "text-to-image"]
status = "active"
role_tags = ["image-flagship", "image-gen"]
fallback = "flux-pro-1-1"
notes = "fal.ai FLUX.2 [pro]. ZERO-CONFIG: no guidance_scale parameter (model handles internally). Per-megapixel billing: $0.03 first MP + $0.015 per extra MP. 1024x1024 = $0.03, 1920x1080 = $0.045. output_per_mtok_micro = 3000000 (base 1MP price). Real provider: fal."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 3000000
status = "verified"
source_url = "https://fal.ai/models/fal-ai/flux-2-pro"
verified_at = "2026-05-01"
[[models]]
id = "flux-pro-1-1"
provider = "local"
display_name = "FLUX.1 [pro] v1.1 Text to Image"
context_tokens = 0
capabilities = ["image-gen", "text-to-image"]
status = "active"
role_tags = ["image-flagship", "image-gen"]
fallback = ""
notes = "fal.ai FLUX Pro 1.1. Per-megapixel billing: $0.04/MP rounded up. Fallback for flux-2-pro. output_per_mtok_micro = 4000000 (1MP base). Real provider: fal."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 4000000
status = "verified"
source_url = "https://fal.ai/models/fal-ai/flux-pro/v1.1"
verified_at = "2026-05-01"
[[models]]
id = "kling-o3"
provider = "local"
display_name = "Kling O3 4K Text/Image to Video"
context_tokens = 0
capabilities = ["video-gen", "text-to-video", "image-to-video"]
status = "active"
role_tags = ["video-gen"]
fallback = ""
notes = "fal.ai Kling O3. 4K video generation. Per-second cost: $0.42/s (audio off). Durations 3-15s; default 5s (=$2.10). 2500-char prompt limit (O3 hard limit). Supports elements + voice_ids simultaneously. output_per_mtok_micro = 42000000 (per-second micro-cents). Real provider: fal."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 42000000
status = "verified"
source_url = "https://fal.ai/models/fal-ai/kling-video/o3/4k/text-to-video"
verified_at = "2026-05-01"
[[models]]
id = "veo-3"
provider = "local"
display_name = "Google Veo 3 (via fal.ai)"
context_tokens = 0
capabilities = ["video-gen", "text-to-video"]
status = "active"
role_tags = ["video-gen"]
fallback = "kling-o3"
notes = "fal.ai Google Veo 3 partner model. Per-second cost: $0.50/s (audio off) or $0.75/s (audio on). Fast tier: $0.25/s (audio off) or $0.40/s (audio on). output_per_mtok_micro = 50000000 (standard audio-off). Real provider: fal/google."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 50000000
status = "verified"
source_url = "https://fal.ai/models/fal-ai/veo3"
verified_at = "2026-05-01"
[[models]]
id = "ideogram-v3"
provider = "local"
display_name = "Ideogram V3 Text to Image"
context_tokens = 0
capabilities = ["image-gen", "text-to-image", "image-edit"]
status = "active"
role_tags = ["image-flagship", "image-gen"]
fallback = ""
notes = "fal.ai Ideogram V3. Per-image tiered: $0.03 TURBO / $0.06 BALANCED / $0.09 QUALITY. Exceptional text/typography rendering. output_per_mtok_micro = 6000000 (BALANCED tier). Real provider: fal."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 6000000
status = "verified"
source_url = "https://fal.ai/models/fal-ai/ideogram/v3"
verified_at = "2026-05-01"
[[models]]
id = "recraft-v3"
provider = "local"
display_name = "Recraft V3 Text to Image / Vector"
context_tokens = 0
capabilities = ["image-gen", "text-to-image", "image-edit"]
status = "active"
role_tags = ["image-cheap", "image-gen"]
fallback = ""
notes = "fal.ai Recraft V3. Per-image: $0.04 (raster) or $0.08 (vector_illustration style). Styles: realistic_image, digital_illustration, vector_illustration. Strong text rendering. output_per_mtok_micro = 4000000 (raster base). Real provider: fal."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 4000000
status = "verified"
source_url = "https://fal.ai/models/fal-ai/recraft-v3"
verified_at = "2026-05-01"
# ============================================================
# ElevenLabs voice generation — verified 2026-05-01
# Source: https://elevenlabs.io/pricing/api
# SCHEMA NOTE: provider="local" placeholder (real provider: elevenlabs).
# Voice pricing: output_per_mtok_micro = per-1k-char micro-cents
# (1 unit = 1000 characters). input_per_mtok_micro = 0.
# ============================================================
[[models]]
id = "elevenlabs-v3"
provider = "local"
display_name = "ElevenLabs Turbo v3 (Flash)"
context_tokens = 0
capabilities = ["voice-gen", "text-to-speech", "voice-clone"]
status = "active"
role_tags = ["voice-gen"]
fallback = "elevenlabs-multilingual-v2"
notes = "ElevenLabs high-quality voice model. Per-1k-chars: $0.10. Low latency ~250-300ms. Supports 32 languages. 40k char limit per request. output_per_mtok_micro = 10000000 (per-1k-chars micro-cents). Real provider: elevenlabs. ElevenLabs 3-step pattern: designVoice -> createVoice -> TTS."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 10000000
status = "verified"
source_url = "https://elevenlabs.io/pricing/api"
verified_at = "2026-05-01"
[[models]]
id = "elevenlabs-multilingual-v2"
provider = "local"
display_name = "ElevenLabs Multilingual v2"
context_tokens = 0
capabilities = ["voice-gen", "text-to-speech", "voice-clone"]
status = "active"
role_tags = ["voice-gen"]
fallback = ""
notes = "ElevenLabs multilingual voice model (fallback for v3). Per-1k-chars: $0.10. Same pricing tier as v3. Supports 32 languages. output_per_mtok_micro = 10000000. Real provider: elevenlabs."
[models.pricing]
input_per_mtok_micro = 0
output_per_mtok_micro = 10000000
status = "verified"
source_url = "https://elevenlabs.io/pricing/api"
verified_at = "2026-05-01"
Reference in a new issue View git blame Copy permalink