mirror of
https://github.com/open-jarvis/OpenJarvis.git
synced 2026-07-30 19:02:16 +00:00
Jon Saad-FalconandClaude Opus 4.6
2aebcd7d77
Trace-driven learning pipeline: - TrainingDataMiner: extract SFT/routing/agent pairs from traces - LoRATrainer: fine-tune local models from trace-derived data - AgentConfigEvolver: rewrite agent configs from trace analysis - LearningOrchestrator: coordinate mine→train→evolve cycle, wired into SystemBuilder Eval framework (15 real IPW benchmarks): - Datasets: SuperGPQA, GPQA, MMLU-Pro, MATH-500, Natural Reasoning, HLE, SimpleQA, WildChat, IPW, GAIA, FRAMES, SWE-bench, SWEfficiency, TerminalBench, TerminalBench Native - Scorers: MCQ extraction, LLM-judge, exact match, structural validation - CLI: jarvis eval list|run|compare|report Composable abstractions: - Recipe system: TOML composition of all 5 pillars (3 built-in recipes) - Agent templates: 15 pre-configured TOML manifests with system prompts - Bundled skills: 20 ready-to-use TOML skill manifests - Operator recipes: researcher (4h), correspondent (5min), sentinel (2h) 102 files changed, ~11,500 lines added. 3241 tests pass (44 skipped). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
11 lines
660 B
TOML
11 lines
660 B
TOML
[template]
|
|
name = "debugger"
|
|
description = "Systematic debugging and root-cause analysis"
|
|
|
|
[agent]
|
|
type = "native_react"
|
|
max_turns = 15
|
|
temperature = 0.2
|
|
tools = ["file_read", "code_interpreter", "shell_exec", "think"]
|
|
system_prompt = """You are a systematic debugger. Given a bug report or error trace, you methodically narrow down the root cause by reading relevant source files, forming hypotheses, and testing them with code execution or shell commands. You follow a disciplined observe-hypothesize-test cycle and avoid jumping to conclusions. When you identify the root cause, you propose a minimal, targeted fix and explain why it resolves the issue."""
|