From 75b5db52f27caae93c9afda908e3733a283c5021 Mon Sep 17 00:00:00 2001 From: Pavel Tisnovsky Date: Mon, 6 Jul 2026 08:56:47 +0200 Subject: [PATCH] LCORE-2861: Sync config doc --- docs/config.html | 16 ++++++++++++++++ docs/config.json | 14 ++++++++++++++ docs/config.md | 14 ++++++++------ 3 files changed, 38 insertions(+), 6 deletions(-) diff --git a/docs/config.html b/docs/config.html index a13e49fb1..29bdff2c3 100644 --- a/docs/config.html +++ b/docs/config.html @@ -1034,6 +1034,22 @@

InferenceConfiguration

default_model / default_provider keep their query-time routing meaning and are independent of this list. + + max_infer_iters + integer + Server-side default for the maximum number of inference iterations a +model can perform in a single request. Prevents small models from +looping indefinitely on tool calls. Per-request values take precedence +over this default. Set to None to disable the limit. + + + max_tool_calls + integer + Server-side default for the maximum number of tool calls allowed in +a single response. Prevents small models from exhausting the context +window with repeated tool calls. Per-request values take precedence over +this default. Set to None to disable the limit. +

JsonPathOperator

diff --git a/docs/config.json b/docs/config.json index ec2bccfa6..01ee5d10b 100644 --- a/docs/config.json +++ b/docs/config.json @@ -896,6 +896,20 @@ }, "title": "High-level inference providers", "type": "array" + }, + "max_infer_iters": { + "type": "integer", + "nullable": true, + "default": 10, + "description": "Server-side default for the maximum number of inference iterations a model can perform in a single request. Prevents small models from looping indefinitely on tool calls. Per-request values take precedence over this default. Set to None to disable the limit.", + "title": "Default max inference iterations" + }, + "max_tool_calls": { + "type": "integer", + "nullable": true, + "default": 30, + "description": "Server-side default for the maximum number of tool calls allowed in a single response. Prevents small models from exhausting the context window with repeated tool calls. Per-request values take precedence over this default. Set to None to disable the limit.", + "title": "Default max tool calls" } }, "title": "InferenceConfiguration", diff --git a/docs/config.md b/docs/config.md index e7ebc5f90..25134a8ee 100644 --- a/docs/config.md +++ b/docs/config.md @@ -332,12 +332,14 @@ In-memory cache configuration. Inference configuration. -| Field | Type | Description | -|------------------|--------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| -| default_model | string | Identification of default model used when no other model is specified. | -| default_provider | string | Identification of default provider used when no other model is specified. | -| context_windows | object | Map of fully-qualified model identifier (e.g., "openai/gpt-4o-mini") to context window size in tokens. Used by the conversation compaction trigger to decide when older turns must be summarized before the input exceeds the window. Models absent from this map have no registered window — callers fall back to their own default or skip the token-based trigger. | -| providers | array | Unified-mode synthesis input (Decision S5): a high-level, backend-agnostic list of inference providers the synthesizer expands into Llama Stack provider entries. Lives at the configuration root so it survives a future backend change. A non-empty list signals unified mode. Empty (the default) leaves legacy/remote modes unaffected. The sibling default_model / default_provider keep their query-time routing meaning and are independent of this list. | +| Field | Type | Description | +|------------------|---------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| default_model | string | Identification of default model used when no other model is specified. | +| default_provider | string | Identification of default provider used when no other model is specified. | +| context_windows | object | Map of fully-qualified model identifier (e.g., "openai/gpt-4o-mini") to context window size in tokens. Used by the conversation compaction trigger to decide when older turns must be summarized before the input exceeds the window. Models absent from this map have no registered window — callers fall back to their own default or skip the token-based trigger. | +| providers | array | Unified-mode synthesis input (Decision S5): a high-level, backend-agnostic list of inference providers the synthesizer expands into Llama Stack provider entries. Lives at the configuration root so it survives a future backend change. A non-empty list signals unified mode. Empty (the default) leaves legacy/remote modes unaffected. The sibling default_model / default_provider keep their query-time routing meaning and are independent of this list. | +| max_infer_iters | integer | Server-side default for the maximum number of inference iterations a model can perform in a single request. Prevents small models from looping indefinitely on tool calls. Per-request values take precedence over this default. Set to None to disable the limit. | +| max_tool_calls | integer | Server-side default for the maximum number of tool calls allowed in a single response. Prevents small models from exhausting the context window with repeated tool calls. Per-request values take precedence over this default. Set to None to disable the limit. | ## JsonPathOperator