feat(prime): continual harness, autonomous budgets, family messaging - #248
Merged
Conversation
Three adoptions from Prime Agent's RLM agent. Continual harness (internal/intelligence/harness): - Versioned, evidence-backed refinement store for supplemental agent state (prompts, memories, skill descriptions, subagent specs). - Refine bumps entry version and records a refinement event (trigger -> changes -> evidence -> outcome); every change is auditable and reversible via snapshot-based rollback. Never rewrites the immutable base system prompt. Durable JSON persistence. Bounded autonomous budgets (internal/engine/autonomous_budget.go): - Tracks turns/tokens/time/continuations within configured limits and a status machine that reports WHY a run stopped: budget_limited:<which> vs gate_passed vs error (a passed gate checks only what it verifies; reaching a limit does not imply success). First budget hit wins and terminal reasons stick. Agent family messaging (internal/multiagent/family_messaging.go): - Direct inter-agent messages scoped to a family graph (parent / sibling / child) rather than global, with per-agent pending caps and token- bucket rate limiting to prevent flooding. Verification: new suites green (harness 8, autonomous 8, family 7); engine + multiagent suites pass; golangci-lint 0 issues; gofmt clean.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Three adoptions from PrimeIntellect-ai/prime-agent (18k★, the RLM self-improving agent).
1. Continual harness —
internal/intelligence/harnessVersioned, evidence-backed refinement of supplemental agent state (prompts/memories/skill descriptions/subagent specs).
Refinebumps the entry version and records a refinement event (trigger → changes → evidence → outcome); every change is auditable and reversible via snapshot-based rollback. Never rewrites the immutable base system prompt. Durable JSON persistence.2. Bounded autonomous budgets —
internal/engine/autonomous_budget.goTracks turns/tokens/time/continuations within configured limits and a status machine that reports why a run stopped:
budget_limited:<which>vsgate_passedvserror. First budget hit wins; terminal reasons stick.3. Agent family messaging —
internal/multiagent/family_messaging.goDirect inter-agent messages scoped to a family graph (parent/sibling/child), with per-agent pending caps and token-bucket rate limiting.
Testing
New suites green (harness 8, autonomous 8, family 7); full engine + multiagent suites pass; golangci-lint 0 issues; gofmt clean.