Redis integration guides¶
The standard way to integrate your application with Redis is to use the official Redis clients and libraries for your programming language. Redis ADBC is another option for workflows that benefit from SQL and Apache Arrow, including the data-tool integrations described here.
This index collects the integration guides linked from the README, including the dedicated DuckDB guide. The shared example setup supports the additional local workflows. The 36 guides with a bidirectional native or custom ADBC path now verify tool outputs written back to Redis as well as reads. See the read/write coverage and rerun commands and 2026-11-01 results.
See the per-guide test results.
The upstream acceptance audit covers all 51 tools in the comparison, with Sling's standalone workflow recorded separately. The runner wires reusable SQLAlchemy, DBI, Fivetran and ADBC contracts into the testing harness, with explicit adapter gaps and skip counts.
Sling has a separate PostgreSQL round-trip test: two 10,003-row round trips passed with zero value or duplicate differences on verified Redis Software 4x QPF. It requires template overrides and direct loading into a precreated Redis table; default staging and empty-result handling have documented failures. These results are separate from the historical Python guide sweep and upstream acceptance counts.
See installation routes for all 51 inventory entries.
Start with installation and diagnostics to choose a package and verify the driver before configuring a tool.
Each guide covers Redis prerequisites, installation, a concrete integration path, and its limits. These are separately maintained driver integrations; availability of a guide does not imply upstream endorsement, release-wheel validation, or an upstream contribution's acceptance.
Support and test evidence distinguishes recorded environments from unverified paths. Upstream contributions are tracked in issue #236 and the existing per-integration issues linked there.
Upstream contribution status records submitted PRs and the prepared changes requiring human or maintainer action.
Workflow provenance¶
PR #248 adds NumPy, scikit-learn, Jupyter notebook, Hugging Face Datasets, Streamlit and SciPy workflows and has merged. Those guides document custom Python data paths and their specific validation limits. Their examples use the public DB-API/Arrow interface; use a checkout containing #248 for the saved workflow runner and tests.
See new workflow documentation status for each upstream project, proposed documentation placement, and submission gaps. Issue #249 tracks the guide contributions and remaining upstream submissions.
Live example validation records Redis Software runs for all eighteen integration paths, including the six new guides, the corrected DuckDB SQL and the dbt Core starter project.
Maintaining the local harness¶
The standard-library integration registry
is the source for guide choices, guide read/write requirements, candidate pytest
selectors, profile membership, Python versions and dependency extras. Update it
when adding an integration; keep dependency pins in
pyproject.toml. Guide names and candidate aliases
are distinct: for example, huggingface-datasets uses candidate datasets,
pyarrow uses arrow, and the R / DBI and RStudio guides share candidate r.
The guide and candidate execution orders are retained separately in the registry.
STANDALONE_GUIDES records integrations with their own external-tool runners;
Sling runs through scripts/test_sling_postgres.py --sling /absolute/path/to/sling
and is not a Python candidate profile or a Python dependency extra.
The runners keep explicit setup hooks for R/Quarto, Kestra and dbt. ray-tls
remains a separate execution mode that owns its TLS fixture. Registry queries
emit plain lines, never shell code. Run the offline contracts with:
python3 -m unittest discover -s scripts -p test_integration_registry.py -v
python3 -m unittest discover -s scripts -p test_candidate_tls_runner.py -v
Real Redis runs still require make -C go redis-up and verified effective
Software 4x QPF before tests; offline runner mocks do not establish integration
compatibility.