reef-infra 0.0.1__tar.gz

reef_infra-0.0.1/.dockerignore

@@ -0,0 +1,24 @@
+.git
+.github
+.pytest_cache
+**/__pycache__
+**/*.pyc
+.claude
+.agents
+imgs
+docs
+*.egg-info
+results
+ckpts
+wandb
+.env
+.venv
+venv
+outputs
+local
+**/*.pkl
+**/*.out
+
+# Example runtime state (reef serve work dirs; may be root-owned from container runs)
+examples/*/work
+examples/*/work-*

reef_infra-0.0.1/.editorconfig

@@ -0,0 +1,20 @@
+# https://editorconfig.org/
+
+root = true
+
+[*]
+charset = utf-8
+end_of_line = lf
+indent_style = space
+indent_size = 4
+trim_trailing_whitespace = true
+insert_final_newline = true
+
+[*.{yaml,yml,json}]
+indent_size = 2
+
+[Makefile]
+indent_style = tab
+
+[*.md]
+trim_trailing_whitespace = false

reef_infra-0.0.1/.github/CODEOWNERS

@@ -0,0 +1,54 @@
+# Default owner for files without a more specific area rule.
+* @BobbyZhouZijian
+
+# Community process, repository automation, packaging, and deployment.
+/.github/ @BobbyZhouZijian
+/docker/ @BobbyZhouZijian
+/pyproject.toml @BobbyZhouZijian
+
+# Core product areas. Keep these entries separate so ownership can be
+# delegated one subsystem at a time as the maintainer group grows.
+/reef/service/ @BobbyZhouZijian
+/reef/scenario/ @BobbyZhouZijian
+/reef/core/ @BobbyZhouZijian
+/recipes/basic/ @BobbyZhouZijian
+/third_party/ @BobbyZhouZijian
+
+# Harness-facing product layer.
+/reef/harness/ @Benjamin-eecs
+/recipes/sao/ @Benjamin-eecs
+/recipes/tttd/ @Benjamin-eecs
+/recipes/openclawrl/ @Benjamin-eecs
+/reef/train/cordis_backend/ @Benjamin-eecs
+/reef/surface/ @Benjamin-eecs
+/third_party/cordis/ @Benjamin-eecs
+/tutorials/harness_evolve/ @Benjamin-eecs
+/recipes/skillclaw/ @Benjamin-eecs
+
+# Runnable examples of the weight methods (the method packages above stay
+# with the product layer).
+/recipes/sao/examples/ @BobbyZhouZijian
+/recipes/tttd/examples/ @BobbyZhouZijian
+/recipes/openclawrl/examples/ @BobbyZhouZijian
+
+# Training execution, artifact handling, and pipeline reliability.
+/reef/train/ @hanfeiyu
+/reef/runtime/ @hanfeiyu
+/reef/artifact/ @hanfeiyu
+/reef/dispatcher.py @hanfeiyu
+/recipes/openclawrl/examples/openclawrl/serve.yaml @hanfeiyu
+
+# Documentation defaults to the Merge Oncall. Domain references override it.
+/docs/ @BobbyZhouZijian
+/docs/user-guide/evolve-your-harness.rst @Benjamin-eecs
+/docs/developer-guide/harness-adapters.rst @Benjamin-eecs
+/docs/developer-guide/write-a-harness-method.rst @Benjamin-eecs
+/docs/developer-guide/write-a-recipe.rst @Benjamin-eecs
+/docs/user-guide/recipes.rst @Benjamin-eecs
+/docs/user-guide/recipes/ @Benjamin-eecs
+/docs/user-guide/evolve-your-model.rst @hanfeiyu
+/docs/developer-guide/loss-families.rst @hanfeiyu
+/docs/reference/cli.rst @BobbyZhouZijian
+# python-api.rst merges the former recipe/surface and batch/processor/preparer
+# pages, so it keeps both prior owners.
+/docs/reference/python-api.rst @Benjamin-eecs @hanfeiyu

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/bug.yml

@@ -0,0 +1,121 @@
+name: Bug report
+description: Report reproducible behavior that is incorrect or unexpected
+title: "[Bug] "
+labels: [bug, "status: needs-triage"]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for reporting a Reef bug. Search existing issues first and
+        remove credentials, private data, provider tokens, and other secrets
+        from every field, log, fixture, and attachment.
+
+  - type: input
+    id: reef_version
+    attributes:
+      label: Reef version
+      description: Provide the installed release, package version, or image tag.
+      placeholder: "For example: 0.7.0 or ghcr.io/...:0.7.0"
+    validations:
+      required: true
+
+  - type: input
+    id: commit
+    attributes:
+      label: Commit
+      description: Provide the full Reef commit hash for a source checkout, or enter N/A for an unmodified release.
+      placeholder: "For example: 0123456789abcdef... or N/A"
+    validations:
+      required: true
+
+  - type: dropdown
+    id: primary_area
+    attributes:
+      label: Primary area
+      description: Choose the area that best owns the failure. Maintainers may refine this during triage.
+      options:
+        - Core and scenario lifecycle
+        - Service, API, or CLI
+        - Training and runtime
+        - Artifact management
+        - Versioning, lineage, or rollback
+        - Observability, metrics, or tracing
+        - Harness, recipes, or surfaces
+        - Documentation
+        - Examples
+        - Packaging or deployment
+        - CI and developer tooling
+        - Not sure
+    validations:
+      required: true
+
+  - type: textarea
+    id: environment
+    attributes:
+      label: Environment
+      description: Include the operating system, Python version, install method, accelerator, and relevant runtime or provider versions.
+      placeholder: |
+        OS: Ubuntu 24.04
+        Python: 3.12.4
+        Install: editable checkout
+        Accelerator: NVIDIA H100, driver ..., CUDA ...
+        Runtime/provider: ...
+    validations:
+      required: true
+
+  - type: textarea
+    id: reproduction
+    attributes:
+      label: Minimal reproduction
+      description: Provide the smallest complete commands, configuration, and code that reproduce the problem.
+      placeholder: |
+        1. Run ...
+        2. Configure ...
+        3. Observe ...
+    validations:
+      required: true
+
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected behavior
+      description: Describe the result you expected and the contract or documentation that led to that expectation.
+    validations:
+      required: true
+
+  - type: textarea
+    id: actual
+    attributes:
+      label: Actual behavior
+      description: Describe what happened instead, including whether the behavior is consistent or intermittent.
+    validations:
+      required: true
+
+  - type: textarea
+    id: logs
+    attributes:
+      label: Logs and diagnostics
+      description: Paste relevant complete errors, stack traces, and logs, or attach a file. Redact all secrets and private data.
+      placeholder: "Paste diagnostics here. Use fenced code blocks for text logs."
+    validations:
+      required: false
+
+  - type: textarea
+    id: additional_context
+    attributes:
+      label: Additional context
+      description: Add screenshots, links, suspected causes, workarounds, or regression information.
+    validations:
+      required: false
+
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Submission checks
+      options:
+        - label: I searched existing issues and did not find the same bug.
+          required: true
+        - label: I reduced the report to a minimal reproduction or explained why that is not possible.
+          required: true
+        - label: I removed credentials, tokens, private data, and other secrets from this report and its attachments.
+          required: true

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,8 @@
+blank_issues_enabled: false
+contact_links:
+  - name: Report a security vulnerability privately
+    url: https://github.com/Human-Agent-Society/reef/blob/main/SECURITY.md
+    about: Read the private reporting instructions; do not open an issue with vulnerability details.
+  - name: Read the contribution guide
+    url: https://github.com/Human-Agent-Society/reef/blob/main/CONTRIBUTING.md
+    about: Check contribution, RFC, testing, and review requirements before starting substantial work.

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/example.yml

@@ -0,0 +1,73 @@
+name: Runnable example
+description: Propose a maintained recipe, demo, or reference deployment
+title: "[Example] "
+labels: ["type: example", "status: needs-triage"]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Examples are maintained user-facing paths, not one-off experiment
+        scripts. Describe the audience, clean-checkout workflow, expected
+        result, validation, and long-term dependency cost.
+
+  - type: dropdown
+    id: primary_area
+    attributes:
+      label: Primary area
+      options:
+        - Core and scenario lifecycle
+        - Service, API, or CLI
+        - Training and runtime
+        - Artifact management
+        - Versioning, lineage, or rollback
+        - Observability, metrics, or tracing
+        - Harness, recipes, or surfaces
+        - Examples
+        - Packaging or deployment
+        - Multiple areas
+    validations:
+      required: true
+
+  - type: textarea
+    id: audience
+    attributes:
+      label: Audience and user journey
+      description: Who will run this example, and what complete workflow should it teach?
+    validations:
+      required: true
+
+  - type: textarea
+    id: command
+    attributes:
+      label: Clean-checkout workflow
+      description: Describe installation, configuration, commands, required services or hardware, and the expected result.
+    validations:
+      required: true
+
+  - type: textarea
+    id: validation
+    attributes:
+      label: Validation and acceptance criteria
+      description: Explain how CI, a smoke lane, or retained evidence will prove that the documented path still works.
+    validations:
+      required: true
+
+  - type: textarea
+    id: maintenance
+    attributes:
+      label: Dependencies and maintenance
+      description: List external projects, models, providers, images, credentials, licenses, and expected ownership.
+    validations:
+      required: true
+
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Submission checks
+      options:
+        - label: I searched existing examples and issues for an equivalent workflow.
+          required: true
+        - label: This is intended as a maintained user-facing path rather than a one-off experiment script.
+          required: true
+        - label: The proposed workflow does not require committing credentials, private data, or sensitive prompts.
+          required: true

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/experiment.yml

@@ -0,0 +1,81 @@
+name: Experiment or reproduction
+description: Propose a benchmark, paper reproduction, or empirical evaluation
+title: "[Experiment] "
+labels: ["type: experiment", "status: needs-triage"]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Define the empirical question and the evidence needed to answer it.
+        Pin models, datasets, prompts, code revisions, and evaluation methods
+        closely enough that another contributor can reproduce the result.
+
+  - type: dropdown
+    id: primary_area
+    attributes:
+      label: Primary area
+      options:
+        - Core and scenario lifecycle
+        - Service, API, or CLI
+        - Training and runtime
+        - Artifact management
+        - Versioning, lineage, or rollback
+        - Observability, metrics, or tracing
+        - Harness, recipes, or surfaces
+        - Examples
+        - CI and developer tooling
+        - Multiple areas
+    validations:
+      required: true
+
+  - type: textarea
+    id: question
+    attributes:
+      label: Question or reproduction target
+      description: State the hypothesis, benchmark claim, or published result to evaluate.
+    validations:
+      required: true
+
+  - type: textarea
+    id: setup
+    attributes:
+      label: Pinned setup
+      description: List model and tokenizer versions, datasets, tasks, prompts, code revisions, seeds, hardware, and runtime configuration.
+    validations:
+      required: true
+
+  - type: textarea
+    id: baselines
+    attributes:
+      label: Baselines and controls
+      description: Identify the comparison runs and the variables that must remain equivalent.
+    validations:
+      required: true
+
+  - type: textarea
+    id: metrics
+    attributes:
+      label: Metrics and acceptance criteria
+      description: Define primary metrics, uncertainty or repeated-run treatment, and the result that would answer the question.
+    validations:
+      required: true
+
+  - type: textarea
+    id: artifacts
+    attributes:
+      label: Reproduction artifacts
+      description: Describe the commands, configs, raw results, curves, logs, and metadata that will be retained. Do not include secrets or private data.
+    validations:
+      required: true
+
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Submission checks
+      options:
+        - label: I searched existing issues, pull requests, RFCs, and experiment results for related work.
+          required: true
+        - label: I separated the empirical question from a preferred implementation.
+          required: true
+        - label: I identified enough pinned inputs and retained evidence for an independent reproduction.
+          required: true

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/feature.yml

@@ -0,0 +1,83 @@
+name: Feature proposal
+description: Propose a focused capability after checking existing issues and RFCs
+title: "[Feature] "
+labels: [enhancement, "status: needs-triage"]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Describe the problem before the implementation. Large public-interface,
+        persistence, trust-boundary, topology, recipe, or training-backend
+        changes may need an RFC before implementation begins.
+
+  - type: dropdown
+    id: primary_area
+    attributes:
+      label: Primary area
+      options:
+        - Core and scenario lifecycle
+        - Service, API, or CLI
+        - Training and runtime
+        - Artifact management
+        - Versioning, lineage, or rollback
+        - Observability, metrics, or tracing
+        - Harness, recipes, or surfaces
+        - Documentation
+        - Examples
+        - Packaging or deployment
+        - CI and developer tooling
+        - Not sure
+    validations:
+      required: true
+
+  - type: textarea
+    id: problem
+    attributes:
+      label: Problem
+      description: Who experiences the problem, and what can they not do today?
+    validations:
+      required: true
+
+  - type: textarea
+    id: use_cases
+    attributes:
+      label: Concrete use cases
+      description: Give representative examples, workloads, or user journeys.
+    validations:
+      required: true
+
+  - type: textarea
+    id: proposed_outcome
+    attributes:
+      label: Proposed outcome
+      description: Describe the externally observable capability, not only an internal implementation.
+    validations:
+      required: true
+
+  - type: textarea
+    id: alternatives
+    attributes:
+      label: Alternatives and workarounds
+      description: What can users do today, and why is it insufficient?
+    validations:
+      required: true
+
+  - type: textarea
+    id: maintenance
+    attributes:
+      label: Compatibility and maintenance
+      description: Describe API, data, deployment, dependency, security, or long-term ownership implications.
+    validations:
+      required: true
+
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Submission checks
+      options:
+        - label: I searched open and closed issues, pull requests, and RFCs for related work.
+          required: true
+        - label: I described the problem and use cases without assuming one implementation.
+          required: true
+        - label: I have not started a large implementation that still needs project-direction approval.
+          required: true

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/performance.yml

@@ -0,0 +1,84 @@
+name: Performance report
+description: Report a reproducible serving, training, harness, or resource regression
+title: "[Performance] "
+labels: [performance, "status: needs-triage"]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Performance reports need a reproducible workload and comparable
+        measurements. Remove credentials, private data, provider tokens, and
+        sensitive prompts from commands, logs, and artifacts.
+
+  - type: dropdown
+    id: primary_area
+    attributes:
+      label: Primary area
+      options:
+        - Core and scenario lifecycle
+        - Service, API, or CLI
+        - Training and runtime
+        - Artifact management
+        - Versioning, lineage, or rollback
+        - Observability, metrics, or tracing
+        - Harness, recipes, or surfaces
+        - Packaging or deployment
+        - Not sure
+    validations:
+      required: true
+
+  - type: textarea
+    id: environment
+    attributes:
+      label: Environment
+      description: Include Reef version/commit, OS, Python, hardware, accelerator stack, runtime, provider, and model.
+    validations:
+      required: true
+
+  - type: textarea
+    id: workload
+    attributes:
+      label: Reproducible workload
+      description: Provide the smallest complete command, configuration, input shape, concurrency, and warm-up procedure.
+    validations:
+      required: true
+
+  - type: textarea
+    id: measurements
+    attributes:
+      label: Measurements
+      description: Include repeated results, units, variance, and the exact measurement method.
+      placeholder: |
+        Baseline: commit ..., 123 requests/s, p95 45 ms
+        Current: commit ..., 98 requests/s, p95 67 ms
+        Runs: 5 after 2 warm-up iterations
+    validations:
+      required: true
+
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected performance or regression boundary
+      description: State the expected result and why the difference is material.
+    validations:
+      required: true
+
+  - type: textarea
+    id: profile
+    attributes:
+      label: Profiles and diagnostics
+      description: Attach sanitized profiles, traces, utilization data, or links to reproducible artifacts.
+    validations:
+      required: false
+
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Submission checks
+      options:
+        - label: I searched open and closed issues for the same performance problem.
+          required: true
+        - label: I compared equivalent configurations and documented all material differences.
+          required: true
+        - label: I removed credentials, private data, and sensitive prompts from all evidence.
+          required: true

reef_infra-0.0.1/.github/ISSUE_TEMPLATE/question.yml

@@ -0,0 +1,58 @@
+name: Usage question
+description: Ask a focused question that the documentation and existing issues do not answer
+title: "[Question] "
+labels: [question, "status: needs-triage"]
+body:
+  - type: dropdown
+    id: primary_area
+    attributes:
+      label: Primary area
+      options:
+        - Core and scenario lifecycle
+        - Service, API, or CLI
+        - Training and runtime
+        - Artifact management
+        - Versioning, lineage, or rollback
+        - Observability, metrics, or tracing
+        - Harness, recipes, or surfaces
+        - Documentation
+        - Examples
+        - Packaging or deployment
+        - CI and developer tooling
+        - Not sure
+    validations:
+      required: true
+
+  - type: textarea
+    id: goal
+    attributes:
+      label: Goal
+      description: What are you trying to accomplish?
+    validations:
+      required: true
+
+  - type: textarea
+    id: attempted
+    attributes:
+      label: What you tried
+      description: Include the relevant documentation, commands, configuration, and observed result.
+    validations:
+      required: true
+
+  - type: textarea
+    id: environment
+    attributes:
+      label: Environment
+      description: Include the Reef version or commit and any relevant runtime, provider, model, OS, or accelerator details.
+    validations:
+      required: true
+
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Submission checks
+      options:
+        - label: I searched the documentation and open and closed issues first.
+          required: true
+        - label: I removed credentials, tokens, private data, and sensitive prompts.
+          required: true