feat(pricing): add custom pricing overrides

[ben-vargas]

Summary

Adds a local custom-pricing.json override layer for exact model pricing fixes. Overrides are loaded from tokscale's config directory, checked before LiteLLM/OpenRouter/Cursor fallback pricing, and use exact matching only so user-specified accounting fixes cannot fuzzy-match into unrelated models.

This is one possible implementation for #548. In the age of agentic development, maintainers may reasonably choose to have their own agents design or implement a different solution if the issue is accepted; this PR is intended as a concrete, tested example of how the feature could work.

What changed

• Add crates/tokscale-core/src/pricing/custom.rs for loading and validating custom-pricing.json.
• Support provider-style per-million fields such as input_cost_per_million_tokens, plus LiteLLM-compatible per-token fields such as input_cost_per_token.
• Check custom overrides with raw exact match first, then existing synthetic /models/ normalization, before falling through to the current pricing resolver.
• Preserve explicit source forcing: --provider litellm and --provider openrouter bypass custom overrides, while --provider custom checks only custom overrides.
• Add tokscale pricing list-overrides with text and JSON output.
• Document custom pricing overrides in the README.

AI disclosure

This PR was AI generated with GPT-5.5 (xhigh) and reviewed iteratively with Opus 4.7 (Max). I manually tested the resulting worktree and pushed the branch from my fork.

Validation

cargo test --workspace
cargo clippy --workspace --all-targets -- -D warnings
cargo build --release --bin tokscale

Also smoke-tested a Fireworks Kimi K2.6 Turbo override:

./target/release/tokscale pricing accounts/fireworks/routers/kimi-k2p6-turbo --provider custom --no-spinner

which reports the custom $2.00 / $8.00 / $0.30 per-1M pricing.

---

Summary by cubic

Add local custom pricing overrides via custom-pricing.json so users can fix or pin model prices. Overrides take precedence over LiteLLM/OpenRouter in lookups and cost calculations, unless a provider is explicitly forced.

• New Features
  • Load exact-match overrides from ~/.config/tokscale/custom-pricing.json (or TOKSCALE_CONFIG_DIR).
  • Supports per-million fields (e.g., input_cost_per_million_tokens) and LiteLLM-style per-token fields (e.g., input_cost_per_token).
  • Lookup order: raw exact match → normalized /models/ match → existing sources (LiteLLM, OpenRouter, Cursor) → fuzzy.
  • Provider forcing preserved: --provider custom checks only custom, --provider litellm|openrouter bypasses custom.
  • Add tokscale pricing list-overrides (text or --json) to inspect loaded overrides.
  • Overrides are loaded once at startup; restart tokscale after editing the file.
  • Docs updated with schema, examples, and usage.

^{Written for commit 1417630. Summary will update on new commits.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
tokscale	Ignored	Preview	May 13, 2026 3:04pm

[cubic-dev-ai]

1 issue found across 4 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="README.md">

<violation number="1" location="README.md:369">
P2: README lookup strategy should document that provider forcing changes custom override precedence</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P2: README lookup strategy should document that provider forcing changes custom override precedence

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At README.md, line 369:

<comment>README lookup strategy should document that provider forcing changes custom override precedence</comment>

<file context>
@@ -357,19 +357,55 @@ tokscale pricing "grok-code"
-5. **Provider Prefix Matching** - Tries common prefixes (`anthropic/`, `openai/`, etc.)
-6. **Cursor Model Pricing** - Hardcoded pricing for models not yet in LiteLLM/OpenRouter (e.g., `gpt-5.3-codex`)
-7. **Fuzzy Matching** - Word-boundary matching for partial model names
+1. **Custom Pricing Overrides** - Exact user-defined entries from `~/.config/tokscale/custom-pricing.json`
+2. **Exact Match** - Direct lookup in LiteLLM/OpenRouter databases
+3. **Alias Resolution** - Resolves friendly names (e.g., `big-pickle` → `glm-4.7`)
</file context>