mirror of
https://github.com/headroomlabs-ai/headroom.git
synced 2026-08-27 14:17:10 -04:00
d5fd84b8df
The documented resolution order was 'built-in defaults -> models.json -> env -> SDK args' with no mention of LiteLLM, which is misleading in a way that costs users work: it reads as though editing models.json is the only way to correct a wrong limit or price, when public models are resolved from LiteLLM and stay current on their own. Restated as an explicit first-match-wins list for both limits and pricing: explicit config -> LiteLLM -> built-in tables -> pattern -> default, noting that step 3 is a fallback for installs without LiteLLM (the dep is gated python_version < '3.14') and models it doesn't know. Also: - Both trees used 'gpt-5: 256000' and 'gpt-5: [5.00, 15.00]' as the example overrides. Both numbers are wrong (272K input, $1.25/$10.00) and gpt-5 is now a known model, so the example taught readers to override a correct value with a stale one. Swapped for a fine-tune and a private deployment -- the cases configuration is actually for. - Noted that limits are an INPUT budget (max_input_tokens), not the total window: gpt-5 is 272K in + 128K out, not 400K of prompt room. - Added gpt-4.1 / gpt-5 / o4 to the pattern-inference table and stated that matching prefers the most specific prefix, so gpt-4.1 is not treated as gpt-4. Docs-only; no code change. |
||
|---|---|---|
| .. | ||
| overrides | ||
| plans | ||
| screenshots | ||
| stylesheets | ||
| agno.md | ||
| api.md | ||
| ARCHITECTURE.md | ||
| benchmarks.md | ||
| ccr.md | ||
| cli.md | ||
| compression.md | ||
| configuration.md | ||
| docker-install.md | ||
| errors.md | ||
| filesystem-contract.md | ||
| getting-started.md | ||
| image-compression.md | ||
| index.md | ||
| integration-guide.md | ||
| langchain.md | ||
| LATENCY_BENCHMARKS.md | ||
| learn.md | ||
| LIMITATIONS.md | ||
| macos-deployment.md | ||
| mcp.md | ||
| memory.md | ||
| metrics.md | ||
| network-diff-capture.md | ||
| persistent-installs.md | ||
| proxy.md | ||
| quickstart.md | ||
| sdk.md | ||
| shared-context.md | ||
| strands.md | ||
| text-compression.md | ||
| transforms.md | ||
| troubleshooting.md | ||
| typescript-sdk.md | ||
| vertex.md | ||