{"id":28493346,"url":"https://github.com/qdrant/search-feedback-loop","last_synced_at":"2026-02-28T13:06:30.908Z","repository":{"id":279313950,"uuid":"935438224","full_name":"qdrant/search-feedback-loop","owner":"qdrant","description":"Using discovery API for improving agentic RAG","archived":false,"fork":false,"pushed_at":"2025-02-19T13:49:18.000Z","size":10,"stargazers_count":3,"open_issues_count":0,"forks_count":0,"subscribers_count":2,"default_branch":"master","last_synced_at":"2025-06-29T10:07:08.608Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","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/qdrant.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}},"created_at":"2025-02-19T12:53:59.000Z","updated_at":"2025-02-20T15:35:03.000Z","dependencies_parsed_at":"2025-02-24T23:43:57.214Z","dependency_job_id":null,"html_url":"https://github.com/qdrant/search-feedback-loop","commit_stats":null,"previous_names":["qdrant/search-feedback-loop"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/qdrant/search-feedback-loop","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/qdrant%2Fsearch-feedback-loop","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/qdrant%2Fsearch-feedback-loop/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/qdrant%2Fsearch-feedback-loop/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/qdrant%2Fsearch-feedback-loop/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/qdrant","download_url":"https://codeload.github.com/qdrant/search-feedback-loop/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/qdrant%2Fsearch-feedback-loop/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":264259914,"owners_count":23580904,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","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":"2025-06-08T09:08:30.512Z","updated_at":"2026-02-28T13:06:25.735Z","avatar_url":"https://github.com/qdrant.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Search Feedback Loop\n## Idea:\nTeach agents to use Discovery API instead of a bad query reformulation.\n## Goal:\nBetter Agentic RAG results.\n\n## Basic Experiment:\n- `all-MiniLM-L6-v2` as main model;\n- `mxbai-embed-large-v1` as agent;\n- BEIR datasets for eval;\n- `precision@1` metric\n  \n### Expensive agent scenario\ntop-10 results of `all-MiniLM-L6-v2` reranked with `mxbai-embed-large-v1`\n\n### Discovery-aware agent scenario\n1. top-3 results of `all-MiniLM-L6-v2` reranked with `mxbai-embed-large-v1`\nIf any results in top-3 changed their order, we have feedback from the agent -- context for discovery\n2. Discovery with `positive context` (top-1 reranked) and `negative context` (top-3 reranked) using `all-MiniLM-L6-v2`, results from 1. excluded\n3. Reranking discovered top-3 with `mxbai-embed-large-v1`\nSelecting the best top-1 result from 1 and 3 based on the `mxbai-embed-large-v1` score.\n\n## How to run\nBEIR datasets folders should be downloaded and put on the same level as scripts;\nIn the current set-up, `Qdrant Cloud` is used, and credentials are taken from `config.ini`.\nSo, to use it also with Cloud, `config.ini` should be changed with your credentials.\n\nThis is the example for running scripts on `FiQa-2018.`\n1. `indexing.py`\n\n```bash\npython indexing.py --dataset_path nfcorpus/corpus.jsonl --total-points-in-dataset 3600 --collection-name \"discovery_agents\"\n```\n2. `evaluating.py`\n\n```bash\npython evaluating.py --input-path-queries nfcorpus/queries.jsonl --input-path-qrels nfcorpus/qrels/test.tsv --collection-name \"discovery_agents\" --total-queries-in-dataset 323 \n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fqdrant%2Fsearch-feedback-loop","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fqdrant%2Fsearch-feedback-loop","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fqdrant%2Fsearch-feedback-loop/lists"}