{"id":50322988,"url":"https://github.com/olesyastorchakprojects/agentic_reasoning_playground","last_synced_at":"2026-05-29T04:01:32.159Z","repository":{"id":352071809,"uuid":"1212755585","full_name":"olesyastorchakprojects/agentic_reasoning_playground","owner":"olesyastorchakprojects","description":"Agentic diagnostic assistant for distributed-system incidents: multi-turn RAG, hypothesis updates, evidence packing, golden evals, and failure-attributed run reports.","archived":false,"fork":false,"pushed_at":"2026-05-20T22:27:11.000Z","size":8933,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-21T04:29:49.282Z","etag":null,"topics":["agentic-workflows","ai-agents","distributed-systems","evals","evaluation-metrics","golden-dataset","incident-diagnosis","llm-evaluation","opentelemetry","rag","rust"],"latest_commit_sha":null,"homepage":"","language":"Rust","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/olesyastorchakprojects.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-04-16T17:44:01.000Z","updated_at":"2026-05-20T22:27:16.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/olesyastorchakprojects/agentic_reasoning_playground","commit_stats":null,"previous_names":["olesyastorchakprojects/agentic_reasoning_playground"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/olesyastorchakprojects/agentic_reasoning_playground","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/olesyastorchakprojects%2Fagentic_reasoning_playground","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/olesyastorchakprojects%2Fagentic_reasoning_playground/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/olesyastorchakprojects%2Fagentic_reasoning_playground/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/olesyastorchakprojects%2Fagentic_reasoning_playground/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/olesyastorchakprojects","download_url":"https://codeload.github.com/olesyastorchakprojects/agentic_reasoning_playground/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/olesyastorchakprojects%2Fagentic_reasoning_playground/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33635961,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-05-29T02:00:06.066Z","response_time":107,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agentic-workflows","ai-agents","distributed-systems","evals","evaluation-metrics","golden-dataset","incident-diagnosis","llm-evaluation","opentelemetry","rag","rust"],"created_at":"2026-05-29T04:01:25.994Z","updated_at":"2026-05-29T04:01:32.151Z","avatar_url":"https://github.com/olesyastorchakprojects.png","language":"Rust","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Distributed Diagnostics Assistant\n\nThis repository is a specification-first diagnostic assistant for distributed systems incidents.\n\nIt is built around a simple idea: instead of producing a one-shot free-form answer over incident documents, the system builds a bounded diagnostic state. It retrieves a leading precedent, keeps competing context in view, adds theory-level explanation, proposes one discriminating check, and updates that state when new observations arrive.\n\nKey entry points:\n\n- [Amazon RDS case study](./Documentation/CASE_STUDY_AMAZON_RDS_READER_STALE_READS.md)\n- [Runtime architecture](./Documentation/ARCHITECTURE.md)\n- [Documentation index](./Documentation/README.md)\n\n## Start Here\n\nThe best way to understand this project is to read the [Amazon RDS case study](./Documentation/CASE_STUDY_AMAZON_RDS_READER_STALE_READS.md).\n\nIt shows one diagnostic run across three iterations: the system starts with competing explanations, accepts new observations, refreshes retrieval, rebuilds prompt context, updates hypothesis confidence, and proposes a more targeted discriminating check.\n\nFor a quick architectural picture, start with [Architecture](./Documentation/ARCHITECTURE.md).\n\n## Why This Repository Matters\n\nThis is not a generic RAG repository.\n\nThe project is interesting because it combines:\n\n- a stateful diagnostic loop modeled as `run -\u003e iteration -\u003e step`\n- a bounded diagnostic response shape instead of unconstrained answer text\n- precedent-guided reasoning with primary precedent, alternative context, and theory evidence\n- continuation as an update to prior diagnostic state rather than a restart\n- a specification-first workflow where contracts define behavior before code and tests\n- a Rust runtime and a Rust eval engine that share runtime-owned types and persisted JSON-blob contracts\n- strong observability with OpenTelemetry, Phoenix-facing semantic spans, and explicit evaluation surfaces\n\n## What The System Does\n\nAt a high level, the runtime:\n\n1. normalizes and structures an initial user-reported incident symptom\n2. decides whether the input is diagnostically sufficient or whether it should ask follow-up questions\n3. retrieves a leading precedent, competing incident context, and theory evidence\n4. assembles a bounded prompt context for generation\n5. returns a structured diagnostic state with hypotheses and one next check\n6. accepts later observations and updates that state across continuation iterations\n\nThe goal of the first response is usually not to claim a final root cause.\nThe goal is to produce the best current diagnostic frame and the next most useful check.\n\n## Core Design Ideas\n\n- `Diagnostic state instead of free-form answers`: the system returns structured hypotheses, competing interpretation, and a discriminating check.\n- `Continuation as state update`: later observations refine the current case instead of restarting the investigation.\n- `Specification-first development`: specs are the source of truth for behavior, tests, and important observability expectations.\n- `Shared runtime/eval contracts`: the Rust eval engine reuses runtime-owned types rather than redefining parallel models.\n- `Evidence-backed repository structure`: implementation, specifications, measurement, evidence, and documentation are intentionally separated.\n- `Observability as a first-class concern`: traces, semantic OpenInference spans, and evaluation outputs are part of the engineering model.\n\n## Repository Structure\n\nThe top-level layout is intentional:\n\n- `Execution/` contains runnable code, including the runtime crate and the eval binary.\n- `Specification/` contains the authoritative behavior and contract definitions.\n- `Measurement/` contains measurement and visualization assets.\n- `Evidence/` contains incident knowledge artifacts and produced evaluation outputs.\n- `Documentation/` contains architecture, design, case-study, and project explanation documents.\n\nFor a fuller walkthrough, see [Documentation/REPOSITORY_MAP.md](./Documentation/REPOSITORY_MAP.md).\n\n## Documentation Entry Points\n\nThe main documentation hub is [Documentation/README.md](./Documentation/README.md).\n\nGood reading order:\n\n1. [Documentation/CASE_STUDY_AMAZON_RDS_READER_STALE_READS.md](./Documentation/CASE_STUDY_AMAZON_RDS_READER_STALE_READS.md)\n2. [Documentation/OVERVIEW.md](./Documentation/OVERVIEW.md)\n3. [Documentation/ARCHITECTURE.md](./Documentation/ARCHITECTURE.md)\n4. [Documentation/KEY_ENGINEERING_DECISIONS.md](./Documentation/KEY_ENGINEERING_DECISIONS.md)\n5. [Documentation/SPECIFICATION_FIRST_APPROACH.md](./Documentation/SPECIFICATION_FIRST_APPROACH.md)\n6. [Documentation/EVALUATION_STORY.md](./Documentation/EVALUATION_STORY.md)\n7. [Documentation/OBSERVABILITY_STORY.md](./Documentation/OBSERVABILITY_STORY.md)\n\n## Specification-First Workflow\n\nOne of the strongest repository-level decisions is that the specification is the source of truth.\n\nIn practice:\n\n- contracts, types, rules, and boundaries are defined in `Specification/`\n- code and tests are generated or implemented against those specs\n- generated or current code is reviewed against the specification, not treated as the design authority\n\nThis is especially important in a repository where runtime behavior, persisted artifacts, eval logic, and observability expectations all need to stay aligned.\n\n## Evaluation And Observability\n\nThe repository includes more than a runtime implementation.\n\nIt also includes:\n\n- a Rust eval engine under `Execution/distributed_diagnostics_eval/`\n- iteration-based evaluation logic and artifacts\n- observability surfaces described in [Documentation/OBSERVABILITY_STORY.md](./Documentation/OBSERVABILITY_STORY.md)\n- a Phoenix-facing OpenInference semantic slice inside the same OTEL trace, rather than a separate parallel trace\n\nThe project is designed to make runtime behavior inspectable, comparable, and reviewable rather than opaque.\n\n## Current Status\n\nThe repository already contains:\n\n- a documented runtime architecture\n- a documented reasoning model\n- a concrete multi-iteration case study\n- a specification-first contract layer\n- an eval story and observability story\n\nDetailed bring-up docs are still being consolidated.\n\n## Getting Started\n\nIf you are new to the repository:\n\n1. start with the [Amazon RDS case study](./Documentation/CASE_STUDY_AMAZON_RDS_READER_STALE_READS.md)\n2. then read [Documentation/README.md](./Documentation/README.md)\n3. use [AGENTS.md](./AGENTS.md) for repository-local working conventions\n4. open `Specification/` when you need the authoritative contracts and runtime rules\n\n## Additional Guides\n\n- [Documentation/README.md](./Documentation/README.md): documentation index\n- [AGENTS.md](./AGENTS.md): repository-local working guide for coding agents\n- [Documentation/REPOSITORY_MAP.md](./Documentation/REPOSITORY_MAP.md): repository layout and navigation\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Folesyastorchakprojects%2Fagentic_reasoning_playground","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Folesyastorchakprojects%2Fagentic_reasoning_playground","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Folesyastorchakprojects%2Fagentic_reasoning_playground/lists"}