feat(tools): add delegation model pool with per-call model/provider override#5229
Closed
HenkDz wants to merge 1 commit into
Closed
feat(tools): add delegation model pool with per-call model/provider override#5229HenkDz wants to merge 1 commit into
HenkDz wants to merge 1 commit into
Conversation
393f343 to
c2ffc7f
Compare
…verride Add intelligent model routing for subagent delegation via a configurable model pool and per-call overrides. - delegation.pool: users define models with strengths, injected into delegate_task schema so the LLM picks the best model per task - model param: per-call or per-task (batch mode) model override - provider param: per-call provider override with full credential resolution - Pool validation: if a model is set but not in pool, falls back with warning - Dynamic schema: _build_delegate_schema() reads pool from config at load time Resolution order (highest to lowest): 1. Per-call model param (top-level, applies to all tasks) 2. Per-task model field (batch mode only) 3. Pool validation — if model set but not in pool, falls back with warning 4. delegation.model from config (fallback default) 5. Parent agent's model (inherited) Supersedes NousResearch#3794 (clean rebase on latest main).
c2ffc7f to
e38deec
Compare
|
Pool + per-call override layering looks clean, and the 5-level resolution hierarchy is explicit. One question: does the dynamic schema builder regenerate the |
Contributor
Author
cached, The only way it would rebuild is if you changed your config and restarted hermes, which is expected behavior anyway. |
MestreY0d4-Uninter
pushed a commit
to MestreY0d4-Uninter/hermes-agent
that referenced
this pull request
Apr 13, 2026
… reasoning_effort Unified implementation combining tier profiles (from stale PR NousResearch#5692) with model pool validation (inspired by PR NousResearch#5229). Features: - 5 named tiers: light, heavy, review, planning, research - Each tier configures model, provider, reasoning_effort, max_iterations - Reasoning floor guardrails prevent silent degradation: heavy/research >= medium, planning/review >= high - Per-task tier in batch mode overrides top-level tier - Optional delegation pool for model validation - override_reasoning_effort in _build_child_agent - resolve_tier_config() merges tier over flat base config - Schema updated with tier enum at top-level and per-task Resolution order: task.tier > top-level tier > default_tier > flat config > parent Config example: delegation: default_tier: heavy tiers: light: {model: gpt-5.4-mini, reasoning_effort: low, max_iterations: 25} review: {model: gpt-5.4, reasoning_effort: xhigh, max_iterations: 60} pool: - model: gpt-5.4, strengths: coding, debugging Tests: - 56 new unit tests (test_delegate_tiers.py) - 7 real integration tests (test_delegate_tiers_real.py) - 128 total delegate tests passing - Backward compatibility verified (flat configs work unchanged)
This was referenced Apr 13, 2026
Closed
MestreY0d4-Uninter
added a commit
to MestreY0d4-Uninter/hermes-agent
that referenced
this pull request
Apr 14, 2026
… reasoning_effort Unified implementation combining tier profiles (from stale PR NousResearch#5692) with model pool validation (inspired by PR NousResearch#5229). Features: - 5 named tiers: light, heavy, review, planning, research - Each tier configures model, provider, reasoning_effort, max_iterations - Reasoning floor guardrails prevent silent degradation: heavy/research >= medium, planning/review >= high - Per-task tier in batch mode overrides top-level tier - Optional delegation pool for model validation - override_reasoning_effort in _build_child_agent - resolve_tier_config() merges tier over flat base config - Schema updated with tier enum at top-level and per-task Resolution order: task.tier > top-level tier > default_tier > flat config > parent Config example: delegation: default_tier: heavy tiers: light: {model: gpt-5.4-mini, reasoning_effort: low, max_iterations: 25} review: {model: gpt-5.4, reasoning_effort: xhigh, max_iterations: 60} pool: - model: gpt-5.4, strengths: coding, debugging Tests: - 56 new unit tests (test_delegate_tiers.py) - 7 real integration tests (test_delegate_tiers_real.py) - 128 total delegate tests passing - Backward compatibility verified (flat configs work unchanged)
10 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Add intelligent model routing for subagent delegation via a configurable model pool and per-call overrides.
Problem: Users with access to multiple models/providers (e.g. via OpenRouter or Z.AI) had no way to control which model a subagent uses at call time.
Solution:
Delegation model pool (
delegation.pool): Users define a list of models with strengths. The pool is injected into thedelegate_tasktool description, so the LLM sees available models and picks the best fit automatically.Per-call overrides:
modelandproviderparams ondelegate_taskfor explicit routing. Per-taskmodelfield in batch mode.Pool validation: If a model is set but not in the pool, falls back to default with a warning.
Changes
tools/delegate_tool.py: Dynamic schema builder, pool validation, per-task model supporthermes_cli/config.py: Addpoolto DEFAULT_CONFIGcli-config.yaml.example: Document pool configurationwebsite/docs/user-guide/configuration.md: Add Delegation Model Pool sectionModel resolution (highest to lowest)
modelparammodelfield (batch mode)delegation.modelfrom configTest plan