- Phase 1 reward: document 3-tier scoring (0.998 classify+issue_type /
0.55 classify-only / 0.002 wrong action); update action example to
include response field with issue_type
- Phase prompt examples: show Phase 1 requiring issue_type in response
- Benchmark: replace single-model table with 6-model comparison table
(GPT-4o, Claude 3.5 Haiku, Qwen2.5-72B, GPT-4o-mini, Llama-3.3-70B,
Mistral-7B) with per-phase breakdown across 30 tickets
- Project structure: update task file comments 5 -> 10 scenarios each
- Inference log example: show issue_type in Phase 1 action JSON
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add OPENENV ASCII art banner to README
Tarun+1: Claude Sonnet 4.6
0c513e6
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Rewrite README to reflect 3-phase RL MDP architecture
Tarun+1: Claude Sonnet 4.6
48850e0
Full documentation covering: 3-phase MDP explanation with state
transition diagram, per-phase reward tables, observation space
evolution across phases, task level scenarios, benchmark scores,
team directory, agent prompt strategy, API reference, project
structure, reward design philosophy, and roadmap.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add team assignment display, update README with full technical docs
Tarun+1: Claude Sonnet 4.6
c27f546
- app.py / server/app.py: add TEAM_DESCRIPTIONS dict; new team_box
Gradio component shows which team a ticket was routed to with
human-readable descriptions (e.g. "[TKT-M001] Logistics Team — ...")
- All UI functions (ui_reset, ui_auto_step, ui_manual_step) updated
to return 6-tuple including team assignment text
- README.md: complete rewrite with technical architecture diagram,
action/observation space tables, reward formula, benchmark scores,
API reference, project structure, and roadmap
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
phase 1 validation v1
Tarun
a23653a
Initial commit with files from Support-Agent repository
