If you want to collect interviews, browser sessions, and documents without sending raw data to the cloud, local AI qualitative analysis can change your workflow. Hearth (v0.7-preview, July 17, 2026) is an open-source, local-first assistant that can record speech, extract text from PDF/DOCX/XLSX, control a real browser, and persist memory on your machine. Paired with Evidano, you can ingest Hearth transcripts and artifacts for thematic, frequency, and cross-segment analysis while keeping data encrypted and private. This post shows a compact workflow UX and research teams can run today: capture with Hearth, prep and QA locally, then run reproducible qualitative analyses in Evidano.
Key Takeaways
Evidano is an AI-powered qualitative data analysis platform that imports local captures, cleans transcripts, enforces privacy, and produces stakeholder-ready visualizations. Hearth v0.7-preview (published July 17, 2026) makes practical local capture possible by recording voice, snapshots, and many document formats on-device. Together, Hearth and Evidano let research teams collect data locally and convert it into auditable thematic analysis while minimizing cloud exposure.
- Hearth v0.7-preview (published July 17, 2026) installs on Windows via a one-click installer and runs from source on macOS/Linux, enabling local-first capture.
- Hearth reads PDF, DOCX, XLSX, PPTX, CSV, JSON, HTML, EPUB, and IPYNB, reducing manual artifact cleaning before import.
- Evidano imports Hearth exports to run automated cleaning, speaker heuristics, PII redaction, coding, and visuals for stakeholder-ready reports.
- A practical seven-step run-book shows you can capture, QA, import, and analyze a small study in about a week.
Fast take: what changed (source)
Hearth published v0.7-preview on July 17, 2026, adding practical local capture features for research teams. Hearth installs on Windows with a one-click installer and runs from source on macOS/Linux, and the project is available on GitHub.
- Why it matters: Hearth can capture voice (faster-whisper), TTS (Kokoro), screen snapshots, and read/write many document formats locally, making it a practical collector for qualitative datasets without default cloud telemetry.
- Payoff for researchers: local capture plus scripted skills produce reproducible, auditable data exports you can import into Evidano for coding, theme extraction, and visualization.
Findings snapshot (quick numbers)
| Item | Value | Note / Implication |
|---|---|---|
| Release | v0.7-preview (published July 17, 2026) | Early but functional: CLI + desktop app are daily drivers; voice/features preview-quality |
| Installer sizes | Full ~1 GB; Lite ~500 MB | Full bundles a llama.cpp server; Lite expects an existing local model |
| Platforms | Windows (installer), macOS/Linux (from source) | Windows easiest for non-dev teams; POSIX support available |
| Requirements | Python 3.11+ (source installs) / optional GPU | Supports local GPU inference or OpenAI-compatible endpoints |
| Formats Hearth reads | PDF, DOCX, XLSX, PPTX, CSV, JSON, HTML, EPUB, IPYNB | Direct ingestion of transcripts & artifacts reduces manual cleaning |
What Hearth does (for research teams)
Hearth is a local operator that listens, reads files, drives a browser, and persists memory on-disk, mapping to three core collector capabilities for qualitative research. Hearth supports speech capture with faster-whisper for continuous listening and mid-sentence barge-in, which is useful for interview capture and rough live transcriptions.
- Speech capture + STT: faster-whisper supports continuous listen with mid-sentence barge-in, handy for interview capture and rough live transcriptions.
- File ingestion & parsers: Hearth extracts text from PDFs, DOCX, XLSX, and large documents (chunked summaries), so raw artifacts can be pre-processed before export.
- Browser & desktop tooling: Hearth can script multi-step web tasks (observe usability flows) and take accessibility-aware snapshots of UI elements instead of pixel-only screenshots.
Hearth also includes researcher-friendly features such as shareable skills (repeatable workflows), background jobs for long scans, and sub-agents that parallelize summarization, and everything stays local unless you opt into a cloud model or web fetch.
Implications for researchers; Local AI qualitative analysis
UX researchers
UX researchers can record moderated sessions on-device, extract timestamps and UI snapshots, and export cleaned transcripts and session artifacts for thematic coding. Use Hearth skills to standardize consent checks and redaction prompts before data leaves the machine.
Policy / field researchers
Policy and field researchers can capture multi-lingual interviews locally and avoid sending raw audio to cloud vendors by default. Researchers can leverage Hearth memory files to save contextual metadata (location, interviewer notes) that travel with transcripts.
Product & market teams
Product and market teams can automate repeatable capture: a Hearth skill can run a scripted guerrilla test, collect artifacts, and package them for analysis, lowering coordination costs for distributed teams.
Do more, faster with Evidano (how to map Hearth outputs to rigorous analysis)
Problem: messy local transcripts → Solution
Evidano imports Hearth-produced STT transcripts and extracted text files to run automated cleaning, speaker diarization heuristics, and prepare a codebook. Import those files directly into Evidano to run automated cleaning, speaker diarization heuristics, and prepare a codebook.
Evidano supports transcription imports, custom dictionaries, and PII redaction, so you can clean locally-captured data and enforce compliance before coding.
Problem: inconsistent codes across studies → Solution
Evidano accepts Hearth annotations or skill outputs exported as CSV/JSON and provides codebook import and AI-assisted coding to ensure consistent thematic labels. Export Hearth annotations or skill outputs as CSV/JSON and load them into Evidano to quantify inter-segment frequencies.
Problem: stakeholder-ready visuals → Solution
Evidano generates word clouds, co-occurrence networks, hierarchical code→subcode visualizations, and cross-segment frequency tables from Hearth-captured artifacts for decks and decision memos. Use Evidano to create stakeholder-ready visuals from local captures.
Security & data policy
Both Hearth and Evidano can run locally: Hearth keeps artifacts on your machine while Evidano encrypts uploads and does not use customer data to train third-party models. For research ethics, always obtain consent for recordings and document retention policies.
Checklist; From Hearth capture to Evidano insight (7 steps)
This seven-step run-book shows how to go from Hearth capture to insights in about a week.
- 1) Install Hearth using the Windows installer or source builds and configure a local model or your preferred endpoint, see GitHub.
- 2) Create a Hearth skill for your interview script that includes a consent prompt and redaction reminders.
- 3) Capture 10 sessions using Hearth's /listen and save per-session artifacts to the workspace.
- 4) Export transcripts and PDFs to a staging folder and run quick QA for speaker labels and timestamps.
- 5) Import the staging folder into Evidano and run thematic and frequency analysis.
- 6) Use Evidano cross-segment analysis to compare personas, regions, or task flows and generate visuals.
- 7) Produce a stakeholder brief and attach verbatim quotes (clickable, source-linked) for transparency.
Wrapping up, next moves
Hearth makes local-first data capture practical, and Evidano turns that local data into rigorous, shareable qualitative analyses without exposing raw content to third-party training. If you care about privacy and reproducibility, this pairing is a fast route from interviews and session recordings to actionable themes and visuals.
- Try Hearth on GitHub (v0.7-preview, July 17, 2026).
- Try Evidano for free to see how to import Hearth exports, run thematic and cross-segment analyses, and produce stakeholder-ready visualizations.
Ethics note: recordings and transcripts are research materials, obtain participant consent and follow your institution's data-handling rules before storage or analysis.
FAQ: Local AI qualitative analysis
What is Hearth v0.7-preview and when was it released?
Hearth v0.7-preview is an open-source local assistant released on July 17, 2026. The release adds local capture features including speech capture, TTS, screen snapshots, and file parsers, and the project is hosted on GitHub.
Can Hearth keep my research data local?
Yes, Hearth keeps artifacts on your machine unless you opt into a cloud model or web fetch. Hearth runs locally and supports on-device speech capture and file parsing, which reduces default cloud telemetry.
What file formats can Hearth read for qualitative datasets?
Hearth can read PDF, DOCX, XLSX, PPTX, CSV, JSON, HTML, EPUB, and IPYNB, enabling direct ingestion of transcripts and artifacts. Direct ingestion reduces manual cleaning before import into analysis tools.
How do I get Hearth exports into Evidano?
You import Hearth transcripts and exported CSV/JSON files into Evidano for cleaning, coding, and visualization. Evidano supports transcription imports, custom dictionaries, PII redaction, and codebook import to standardize analysis workflows.