chore: release 0.6.0

deepgym/CHANGELOG.md

@@ -1,7 +1,10 @@
 # Changelog
 
-## Unreleased
+## 0.6.0 (2026-10-23)
 
+- **Agent harnesses:** `deepgym.agents` adds `AnthropicComputerAgent`, `OpenAIComputerAgent`, `ClaudeCodeAgent` and `CodexAgent`, each driving a live session. `DeepGym.run_episode(env, agent, budget=...)` runs the episode and returns `(AgentResult, RunResult)`; the callable interface still returns `(Trajectory, RunResult)`. The Claude Code and Codex CLIs are installed at pinned versions and checked in the sandbox. See `docs/agents.md`.
+- **Computer agents run uncapped (#98):** `AnthropicComputerAgent` streams with `max_tokens=128000` (the model maximum), `OpenAIComputerAgent` sends no `max_output_tokens`, and reasoning effort defaults to `max`. A response that doesn't complete returns `error` and applies no action. `Budget.max_turns`, `Budget.max_seconds` and the environments' `max_steps` accept `None` for no limit. `MultiTurnEnvironment.volumes` mounts Daytona volumes (`VolumeMount`).
+- **API tool agents (#99):** `AnthropicToolAgent` and `OpenAIToolAgent` act only through HTTP routes an environment declares in `tools` (`HttpTool`, `HttpRoute`). Every write stays unsettled until the environment's `root_tool_settle` hook proves the app finished or dropped it; without that proof the result has no verdict (`unsettled_write`). Write tools can't be combined with step verifiers. Tool credentials stay root-only and are redacted from every record. Tool calls run only after a complete response, and cut-off or unparsable responses are recorded and never acted on.
 - **OSWorld computer-use tasks:** `scripts/import_osworld.py` imports 16 of OSWorld's (Apache-2.0) LibreOffice Calc and Writer tasks as computer-use environments, each with a reference episode, and writes OSWorld's license and a NOTICE next to them. They boot the Daytona snapshot `deepgym-osworld-libreoffice` (`--build-snapshot` creates it), since Daytona's default image has no LibreOffice; `metadata.json` takes a `snapshot`. The verifier scores the saved file with the task's own `compare_table` rules (ported into the shared `_checks.py`, standard library only) or its reference text, never a screenshot. Rules the starting file already holds, or holds once an empty sheet is added, are guards, so an untouched file or a blank sheet scores 0; a saved file the readers can't read (missing, corrupt, not a regular file, too large) scores 0, and reading and scoring run in a time- and memory-limited child whose failure also scores 0, so nothing in the file can crash or stall the verifier. `deepgym check` asks a computer-use folder for `reference_actions.json` instead of `reference_solution.*`, and is inconclusive on one off Daytona.
 
 ## 0.5.0 (2026-10-14)

deepgym/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "deepgym"
-version = "0.5.0"
+version = "0.6.0"
 description = "RL training environments with verifiable rewards for coding and computer-use agents"
 readme = "README.md"
 license = {text = "PolyForm-Noncommercial-1.0.0"}

deepgym/src/deepgym/__init__.py

@@ -1,6 +1,6 @@
 """DeepGym — Managed RL training and evaluation infrastructure."""
 
-__version__ = '0.5.0'
+__version__ = '0.6.0'
 
 from deepgym.async_core import AsyncDeepGym
 from deepgym.benchmark_envs import (