feat: add depth>1 recursive subcalls with limits and cost tracking#84
Open
rawwerks wants to merge 5 commits intoalexzhang13:mainfrom
Open
feat: add depth>1 recursive subcalls with limits and cost tracking#84rawwerks wants to merge 5 commits intoalexzhang13:mainfrom
rawwerks wants to merge 5 commits intoalexzhang13:mainfrom
Conversation
- Add max_budget parameter to RLM (in USD) - Add BudgetExceededError exception - Track cumulative cost across iterations - Propagate remaining budget to child RLMs - Add print_budget_exceeded to VerbosePrinter - Export BudgetExceededError from rlm module Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This was referenced Feb 11, 2026
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
This PR adds full depth>1 recursive subcall support for RLM, enabling nested RLM calls when code inside the REPL invokes
llm_query().Features
_subcall()spawns child RLM instances with incremented depth_subcall()returnsRLMChatCompletion(aligns with LM completion struct)max_budgetparameter withBudgetExceededError, OpenRouter cost extractionmax_timeout,max_tokens,max_errorswith corresponding exceptionson_subcall_start,on_subcall_complete,on_iteration_start,on_iteration_completeFiles Changed
rlm/core/rlm.py- Core subcall logic, limits, callbacksrlm/core/types.py- UsageSummary cost aggregationrlm/environments/local_repl.py- subcall_fn integrationrlm/clients/openai.py- OpenRouter cost extractionrlm/logger/verbose.py- Limit exceeded displaydocs/getting-started.md,docs/api/rlm.md- DocumentationTests
google/gemini-3-flash-previewvia OpenRouterBreaking Changes
None - all existing APIs remain compatible.
🤖 Generated with Claude Code