github/openclaw
1
0
Fork 0
mirror of https://github.com/openclaw/openclaw.git synced 2026-05-01 09:36:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(ollama): unify context window handling across discovery, merge, and OpenAI-compat transport (#29205)

Browse Source 
* fix(ollama): inject num_ctx for OpenAI-compatible transport

* fix(ollama): discover per-model context and preserve higher limits

* fix(agents): prefer matching provider model for fallback limits

* fix(types): require numeric token limits in provider model merge

* fix(types): accept unknown payload in ollama num_ctx wrapper

* fix(types): simplify ollama settled-result extraction

* config(models): add provider flag for Ollama OpenAI num_ctx injection

* config(schema): allow provider num_ctx injection flag

* config(labels): label provider num_ctx injection flag

* config(help): document provider num_ctx injection flag

* agents(ollama): gate OpenAI num_ctx injection with provider config

* tests(ollama): cover provider num_ctx injection flag behavior

* docs(config): list provider num_ctx injection option

* docs(ollama): document OpenAI num_ctx injection toggle

* docs(config): clarify merge token-limit precedence

* config(help): note merge uses higher model token limits

* fix(ollama): cap /api/show discovery concurrency

* fix(ollama): restrict num_ctx injection to OpenAI compat

* tests(ollama): cover ipv6 and compat num_ctx gating

* fix(ollama): detect remote compat endpoints for ollama-labeled providers

* fix(ollama): cap per-model /api/show lookups to bound discovery load
This commit is contained in:
Vincent Koc
2026-02-27 17:20:47 -08:00
committed by GitHub
parent 70a4f25ab1
commit f16ecd1dac
14 changed files with 582 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/config/schema.help.ts
View File

@@ -630,7 +630,7 @@ export const FIELD_HELP: Record<string, string> = {
models:
"Model catalog root for provider definitions, merge/replace behavior, and optional Bedrock discovery integration. Keep provider definitions explicit and validated before relying on production failover paths.",
"models.mode":
'Controls provider catalog behavior: "merge" keeps built-ins and overlays your custom providers, while "replace" uses only your configured providers. In "merge", matching provider IDs preserve non-empty agent models.json apiKey/baseUrl values and fall back to config when agent values are empty or missing.',
'Controls provider catalog behavior: "merge" keeps built-ins and overlays your custom providers, while "replace" uses only your configured providers. In "merge", matching provider IDs preserve non-empty agent models.json apiKey/baseUrl values and fall back to config when agent values are empty or missing; matching model contextWindow/maxTokens use the higher value between explicit and implicit entries.',
"models.providers":
"Provider map keyed by provider ID containing connection/auth settings and concrete model definitions. Use stable provider keys so references from agents and tooling remain portable across environments.",
"models.providers.*.baseUrl":
@@ -641,6 +641,8 @@ export const FIELD_HELP: Record<string, string> = {
'Selects provider auth style: "api-key" for API key auth, "token" for bearer token auth, "oauth" for OAuth credentials, and "aws-sdk" for AWS credential resolution. Match this to your provider requirements.',
"models.providers.*.api":
"Provider API adapter selection controlling request/response compatibility handling for model calls. Use the adapter that matches your upstream provider protocol to avoid feature mismatch.",
"models.providers.*.injectNumCtxForOpenAICompat":
"Controls whether OpenClaw injects `options.num_ctx` for Ollama providers configured with the OpenAI-compatible adapter (`openai-completions`). Default is true. Set false only if your proxy/upstream rejects unknown `options` payload fields.",
"models.providers.*.headers":
"Static HTTP headers merged into provider requests for tenant routing, proxy auth, or custom gateway requirements. Use this sparingly and keep sensitive header values in secrets.",
"models.providers.*.authHeader":