strands-sglang 0.1.0__tar.gz

strands_sglang-0.1.0/.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,38 @@
+---
+name: Bug report
+about: Create a report to help us improve
+title: ''
+labels: ''
+assignees: ''
+
+---
+
+**Describe the bug**
+A clear and concise description of what the bug is.
+
+**To Reproduce**
+Steps to reproduce the behavior:
+1. Go to '...'
+2. Click on '....'
+3. Scroll down to '....'
+4. See error
+
+**Expected behavior**
+A clear and concise description of what you expected to happen.
+
+**Screenshots**
+If applicable, add screenshots to help explain your problem.
+
+**Desktop (please complete the following information):**
+ - OS: [e.g. iOS]
+ - Browser [e.g. chrome, safari]
+ - Version [e.g. 22]
+
+**Smartphone (please complete the following information):**
+ - Device: [e.g. iPhone6]
+ - OS: [e.g. iOS8.1]
+ - Browser [e.g. stock browser, safari]
+ - Version [e.g. 22]
+
+**Additional context**
+Add any other context about the problem here.

strands_sglang-0.1.0/.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,20 @@
+---
+name: Feature request
+about: Suggest an idea for this project
+title: ''
+labels: ''
+assignees: ''
+
+---
+
+**Is your feature request related to a problem? Please describe.**
+A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
+
+**Describe the solution you'd like**
+A clear and concise description of what you want to happen.
+
+**Describe alternatives you've considered**
+A clear and concise description of any alternative solutions or features you've considered.
+
+**Additional context**
+Add any other context or screenshots about the feature request here.

strands_sglang-0.1.0/.github/workflows/publish.yml

@@ -0,0 +1,50 @@
+name: Publish to PyPI
+
+on:
+  release:
+    types: [published]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install build tools
+        run: pip install build twine
+
+      - name: Build package
+        run: python -m build
+
+      - name: Check package
+        run: twine check dist/*
+
+      - name: Upload build artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  publish-pypi:
+    needs: build
+    runs-on: ubuntu-latest
+    environment: pypi
+    permissions:
+      id-token: write  # Required for trusted publishing
+
+    steps:
+      - name: Download build artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+      - name: Publish to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        # Uses trusted publishing (OIDC) - no API token needed
+        # Configure at: https://pypi.org/manage/project/strands-sglang/settings/publishing/

strands_sglang-0.1.0/.github/workflows/test.yml

@@ -0,0 +1,65 @@
+name: Tests
+
+on:
+  push:
+    branches: [main, mainline]
+  pull_request:
+    branches: [main, mainline]
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install ruff
+        run: pip install ruff
+
+      - name: Run ruff check
+        run: ruff check src/
+
+      - name: Run ruff format check
+        run: ruff format --check src/
+
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12"]
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Cache pip dependencies
+        uses: actions/cache@v4
+        with:
+          path: ~/.cache/pip
+          key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ hashFiles('pyproject.toml') }}
+          restore-keys: |
+            ${{ runner.os }}-pip-${{ matrix.python-version }}-
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e ".[dev]"
+
+      - name: Run unit tests
+        run: pytest tests/unit/ -v --cov=src/strands_sglang --cov-report=xml
+
+      - name: Upload coverage
+        uses: codecov/codecov-action@v4
+        if: matrix.python-version == '3.12'
+        with:
+          files: ./coverage.xml
+          fail_ci_if_error: false

strands_sglang-0.1.0/.gitignore

@@ -0,0 +1,221 @@
+# Created by https://www.toptal.com/developers/gitignore/api/python,jupyternotebooks
+# Edit at https://www.toptal.com/developers/gitignore?templates=python,jupyternotebooks
+
+### JupyterNotebooks ###
+# gitignore template for Jupyter Notebooks
+# website: http://jupyter.org/
+
+.ipynb_checkpoints
+*/.ipynb_checkpoints/*
+
+# IPython
+profile_default/
+ipython_config.py
+
+# Remove previous ipynb_checkpoints
+#   git rm -r .ipynb_checkpoints/
+
+### Python ###
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+
+# IPython
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+### Python Patch ###
+# Poetry local configuration file - https://python-poetry.org/docs/configuration/#local-configuration
+poetry.toml
+
+# ruff
+.ruff_cache/
+
+# LSP config files
+pyrightconfig.json
+
+# End of https://www.toptal.com/developers/gitignore/api/python,jupyternotebooks
+
+# Ignore uv lock files for now (uncomment if later UV is used)
+uv.lock
+
+# Ignore nohup.out
+nohup.out
+
+# Ignore data downloaded from AWS S3
+# data*/
+package_data_source/
+
+# Ignore peru files
+.hatch/
+# Ignore PeruHatch local cache directory
+/.hatch
+
+# Ignore private directory
+private/
+
+# Ignore temp directory
+temp/
+
+# Ignore test notebooks
+test*.ipynb
+testing/
+
+# Ignore uv.lock
+uv.lock
+
+# Ignore data folder
+/data
+
+# Ignore DS_Store
+.DS_Store

strands_sglang-0.1.0/.pre-commit-config.yaml

@@ -0,0 +1,37 @@
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.6.4
+    hooks:
+      - id: ruff
+        args: [--fix]
+        files: ^(src/strands_sglang/|scripts/|dev/).*\.py$
+      - id: ruff-format
+        files: ^(src/strands_sglang/|scripts/|dev/).*\.py$
+
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.5.0
+    hooks:
+      - id: check-yaml
+      - id: check-case-conflict
+      - id: detect-private-key
+      - id: check-added-large-files
+        args: ['--maxkb=1000']
+      - id: requirements-txt-fixer
+
+  - repo: https://github.com/compilerla/conventional-pre-commit
+    rev: v3.4.0
+    hooks:
+      - id: conventional-pre-commit
+        stages: [commit-msg]
+        args:
+          - feat
+          - fix
+          - docs
+          - style
+          - refactor
+          - perf
+          - test
+          - build
+          - ci
+          - chore
+          - revert

strands_sglang-0.1.0/CHANGELOG.md

@@ -0,0 +1,115 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+## [0.1.0] - 2026-01-20
+
+First beta release. The library is now considered stable for production use in agentic RL training.
+
+### Added
+
+- **Tool Parse Error Tracking**: `SGLangModel` now tracks tool call parse errors via the `tool_parse_errors` attribute. This enables distinguishing between parse errors (malformed JSON, missing tool name) and execution errors (tool threw exception) during RL training metrics collection.
+
+  ```python
+  model = SGLangModel(tokenizer=tokenizer)
+  # After generation:
+  print(model.tool_parse_errors)  # {"tool_name": count, ...}
+  ```
+
+### Changed
+
+- **Beta Status**: Upgraded from Alpha to Beta. The API is now stable and suitable for production RL training workloads.
+
+## [0.0.3] - 2026-01-08
+
+### Added
+
+- **Qwen3 Hybrid Thinking Mode Support**: Added `enable_thinking` config option for Qwen3 hybrid thinking models. This is passed to `apply_chat_template` to control whether the model uses its internal reasoning mode with `<think>` tokens. Default is `None` (not passed to template) to avoid affecting non-Qwen3 models.
+
+  ```python
+  # For Qwen3 hybrid models: enable thinking mode
+  model = SGLangModel(tokenizer=tokenizer, enable_thinking=True)
+
+  # For Qwen3 hybrid models: disable thinking for faster non-reasoning tasks
+  model = SGLangModel(tokenizer=tokenizer, enable_thinking=False)
+
+  # For non-Qwen3 models: don't set (default None, parameter not passed)
+  model = SGLangModel(tokenizer=tokenizer)
+  ```
+
+- **Related Projects**: Added [strands-vllm](https://github.com/agents-community/strands-vllm) to README as a community vLLM provider inspired by this project.
+
+### Changed
+
+- **Simplified TokenManager API**: Removed `tokenizer` parameter from `TokenManager.__init__()` and removed the `decode()` method. The tokenizer was only used for single-token decoding which was never used in practice—all real usage calls `model.tokenizer.decode()` directly for batch decoding with options like `skip_special_tokens`.
+
+- **Message Formatting Methods**: Converted `_format_message_content` and `format_request_messages` to `@classmethod` since they don't use instance state. This clarifies intent and allows calling without an instance.
+
+### Fixed
+
+- **SLIME-Aligned Retry for Local Servers**: Changed retry behavior to match SLIME's aggressive retry philosophy for local SGLang servers during RL training:
+  - 400 errors are now **retried** (can be transient during weight reloading, memory pressure)
+  - Only truly non-retryable: 401 (auth), 403 (forbidden), 404 (not found)
+  - 400 with context length patterns still not retried (won't help)
+  - References: [OpenAI Python SDK](https://github.com/openai/openai-python) retries 408/409/429/5xx; [SLIME](https://github.com/THUDM/slime) retries ALL errors
+
+- **Improved Context Length Detection**: Expanded patterns to detect context/prompt length errors in 400 responses:
+  - Now matches: "exceed", "too long", "max model len", "maximum length", "context length"
+  - These are converted to `ContextWindowOverflowException` (TRUNCATED, not ABORTED)
+  - Added logging for unexpected 400 errors to aid debugging
+
+## [0.0.2] - 2026-01-07
+
+### Added
+
+- **`SGLangClient` Class** (`client.py`): High-level async HTTP client for SGLang server, aligned with [Slime's http_utils.py](https://github.com/THUDM/slime/blob/main/slime/utils/http_utils.py) for RL training stability:
+  - Connection pooling (default 1000 max connections, with matching keepalive)
+  - Aggressive retry: 60 attempts with 1s delay (like Slime)
+  - Infinite timeout by default for long generations (`timeout=None`)
+  - Non-streaming POST for better parallelism at scale
+
+- **`SGLangClient.from_slime_args()` Factory Method**: Create client directly from Slime training args with auto-computed `max_connections`:
+
+  ```python
+  client = SGLangClient.from_slime_args(args)
+  model = SGLangModel(tokenizer=tokenizer, client=client)
+  ```
+
+- **Slime-Aligned Retry Logic**: Aggressive retry on most errors (blacklist approach):
+  - Retries all 5xx server errors
+  - Retries 408 (Request Timeout) and 429 (Rate Limit) per OpenAI best practices
+  - Retries connection errors (`ConnectError`, `PoolTimeout`, `ReadTimeout`)
+  - Only non-retryable: permanent client errors (400, 401, 403, 404, 422, etc.)
+
+- **Conventional Commits**: Added `commit-msg` hook via `conventional-pre-commit` to enforce [Conventional Commits](https://www.conventionalcommits.org/) format.
+
+### Changed
+
+- **Default Port**: Changed from 8000 to 30000 to match SGLang's default.
+- **`SGLangModel` Now Uses `SGLangClient`**: The model uses `SGLangClient` for HTTP communication, providing retry logic and better error handling.
+- **Improved Error Handling**: SGLang HTTP errors now properly raise `ContextWindowOverflowException` for context length errors and `ModelThrottledException` for rate limiting (429/503).
+- **BREAKING: Non-Streaming Only**: `SGLangClient.generate()` now returns `dict[str, Any]` directly instead of an `AsyncGenerator`. This provides ~20x better parallelism for RL training at scale by releasing connections immediately after response.
+
+### Removed
+
+- **Streaming Support**: Removed all streaming/SSE code. Non-streaming POST is now the only mode, aligned with Slime's `http_utils.py` for optimal RL training performance. Streaming held connections open during generation, causing serialization at high concurrency.
+- **`stream` Config Option**: Removed from `SGLangConfig` as streaming is no longer supported.
+
+### Fixed
+
+- Default `max_new_tokens` increased for thinking models that require longer outputs.
+
+## [0.0.1] - 2026-01-03
+
+### Added
+
+- Initial release with SGLang native `/generate` API support.
+- Token-In/Token-Out (TITO) tracking via `TokenManager`.
+- Hermes/Qwen tool call parsing with `HermesToolCallParser`.
+- `ToolIterationLimiter` hook for clean trajectory truncation.
+- Integration with Strands Agents SDK.