This document details the functional features, mathematical mutators, autonomous agents, and scheduling mechanisms implemented in Project Hermit (v0.1.9).
- Purpose: Allows dynamic mutation and optimization of target python skills without modifying or restarting the host orchestrator process.
- Inputs: Target skill name, mutated python code candidate string, frozen harness baseline hash (
harness_hash). - Outputs: Verification test result, updated skill registry entry, hot-reloaded module state (
importlib.reload). - Dependencies:
orchestrator.py,dynamic_mcp_server.py,sqlite3. - Current Maturity: Production Ready (v0.1.9).
- Future Potential: Support remote MCP payloads over standard JSON-RPC HTTP transport.
- Purpose: Prevents queue starvation and single-skill priority monopolization by alternating between exploration of untouched functions and exploitation of performance bottlenecks.
- Inputs:
active_skillsdatabase metrics (latency, complexity, merge counts, consecutive failure counts). - Outputs: Target skill selected for next optimization cycle with designated strategy mode:
- Phase 1: Target untouched skills (0 merges) sorted by optimization headroom (
latency × complexity). - Phase 2: Target high-latency bottleneck skills with a maximum consecutive attempts cap (
MAX_CONSECUTIVE = 3).
- Phase 1: Target untouched skills (0 merges) sorted by optimization headroom (
- Dependencies:
hermit_daemon.py,hermit_memory.db. - Current Maturity: Production Ready (v0.1.9).
- Future Potential: Integrate dynamic reinforcement learning for adaptive phase ratio switching.
- Purpose: Executes untrusted or candidate python code in an isolated 32MB
tmpfsRAM disk with strict virtual memory, CPU time, and AST safety bounds. - Inputs: Candidate code file path, test parameters, CPU time limit (30s), memory bound (
RLIMIT_AS). - Outputs:
SandboxResultobject containing execution exit code, duration (ms), exact grandchild process peak RSS (KB), stdout, and stderr. - Dependencies:
sandbox.py,resource,psutil,ast. - Current Maturity: Production Ready (v0.1.9).
- Future Potential: Add containerized Docker or PRoot sandbox fallback layer for non-Linux hosts.
- Purpose: Applies deterministic mathematical AST transformations to optimize numeric computation routines.
- Inputs: Python source code AST containing numeric or algebraic logic.
- Outputs: Transformed AST with constant folding, expression simplification, and closed-form math substitutions.
- Dependencies:
math_mutator.py,ast. - Current Maturity: Production Ready (Tested via
test_math_mutator.py&test_math_integration.py).
-
Purpose: Replaces
$O(N^2)$ global quadratic energy updates with$O(N)$ local bitwise spin-flip delta calculations. - Inputs: Quadratic Unconstrained Binary Optimization (QUBO) matrix logic and binary vector routines.
- Outputs: Optimized bitwise state update routines.
-
Dependencies:
qubo_mutator.py,ast. -
Current Maturity: Production Ready (Tested via
test_qubo_mutator.py&test_qubo_integration.py).
| Skill Name | Version / Merges | Primary Functionality | Baseline Status |
|---|---|---|---|
hex_search |
v75 | Binary pattern search and memory buffer scanning | Verified & Benchmark Frozen |
scan_allowlist |
v52 | IP / hash allowlist verification and matching | Verified & Benchmark Frozen |
parse_ip_port |
v37 | Socket address parsing and validation | Verified & Benchmark Frozen |
eval_cond |
v21 | Rule evaluation and conditional AST dispatching | Verified & Benchmark Frozen |
generate_hexdump |
v7 | Binary data formatting and string rendering | Verified |
search_disk_timeline |
v2 | Regex-compiled file timeline log scanning | Verified |
analyze_thoughts |
v2 | Token estimation and bitwise text analytics | Verified |