{"id":35299414,"url":"https://github.com/evalece/rag","last_synced_at":"2026-04-08T17:31:54.383Z","repository":{"id":290475368,"uuid":"974572279","full_name":"evalece/RAG","owner":"evalece","description":"My first implementation of Retrieval Augmented Language Models + learning logs ","archived":false,"fork":false,"pushed_at":"2025-06-10T18:53:42.000Z","size":53,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-06-10T19:58:44.678Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/evalece.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"license.txt","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}},"created_at":"2025-04-29T01:40:52.000Z","updated_at":"2025-05-31T23:58:47.000Z","dependencies_parsed_at":"2025-06-10T19:54:19.180Z","dependency_job_id":null,"html_url":"https://github.com/evalece/RAG","commit_stats":null,"previous_names":["evalece/rag"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/evalece/RAG","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/evalece%2FRAG","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/evalece%2FRAG/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/evalece%2FRAG/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/evalece%2FRAG/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/evalece","download_url":"https://codeload.github.com/evalece/RAG/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/evalece%2FRAG/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31566784,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-08T14:31:17.711Z","status":"ssl_error","status_checked_at":"2026-04-08T14:31:17.202Z","response_time":54,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.5:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"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":[],"created_at":"2025-12-30T16:49:18.775Z","updated_at":"2026-04-08T17:31:54.366Z","avatar_url":"https://github.com/evalece.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# RAG \n![Last Commit](https://img.shields.io/github/last-commit/evalece/RAG)\n- [x] In-Development \n\n# May 24 - Readings and Pivots on this project\n- I read (actually, quickly scan through): ReAct, Facilitating Multu-Trun Function Calling For LLMs Via Compositional Institutution Tuning, LoRA, Understanding Zero-Shot and Few-Shot in LLMs\n- Thought: This project can focus on making LLM agent microservice-like:\n    - 1. Configurable in and outside of pipeline \n    - 2. Runnable, test-able in and outside of pipeline \n    - 3. Configuration allows for features mention in several papers (i.e., prompt, LLM, Transformer or other training modules, memory, few or zero-shot).\n    - 4. Monitoring, re-configuration, re-arrangement and benchmarking easy\n    - 5. (I haven't read this part) Considers GPU and other hardware optimization\n\n\u003c!-- ROADMAP --\u003e\n## Roadmap\n- [x] Define Project Roadmap \n- [x] Tool exploration \n    - [x] LLM model test \n    - [ ] Embedding \n    - [ ] Vector DB\n    - [ ] Benchmark \u0026 Baseline testing\n    - [ ] Test method \n- [ ] Simplified RAG test\n- [ ] RAG + LLM benchmark as baseline \n    - [ ] Frozen RAG\n    - [ ] Non-Frozen RAG + Encoder on input and document \n- [ ] Decoder into RAG to train encoders. \n\n\n## Prototype Plan\n\n- A possible prototype: \n    - Speculative RAG \n        - Decoupled decoder in non-training usage for low latency. \n        - Implement Semantic Cache, reference GPTCache [1] \n        - Similar Approach: REALM by Google [1]\n    - Observation of RAG during training and testing \n        - Learning off means model do not learn from the operation for retrieval \n        -  (Learning off on generation+document sets*) Logit RAG in pipeline to produce model confidence on each retrieval and response. \n        - Store learning + toggle feature to run the model with and without learning on \n            - retrieval \n            - decoder backprop (for later) \n        - Stat tracking (i.e., recall, precision, confidence etc )\n    \n*In Logit-Based RAG, generator and retrieval output gets map to a softmax tensor, representing probabilities. In our case, we do not wish to combine generator output to train RAG. -- check if confidence LLM exists, if not, calculating it.\n\n\n## Reference\n  [1] D. Bahdanau, K. Cho, and Y. Bengio, “Neural machine translation by jointly learning to align and translate,” arXiv.org, Sep. 01, 2014. https://arxiv.org/abs/1409.0473\n\n  [2] P. Zhao et al., “Retrieval-Augmented Generation for AI-Generated Content: a survey,” arXiv.org, Feb. 29, 2024. https://arxiv.org/abs/2402.19473\n## Contributor\n\n\u003ca href=\"https://github.com/evalece/RAG/graphs/contributors\"\u003e\n  \u003cimg src=\"https://contrib.rocks/image?repo=evalece/RAG\" /\u003e\n\u003c/a\u003e","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fevalece%2Frag","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fevalece%2Frag","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fevalece%2Frag/lists"}