Skip to content

Releases

FollowThrough v0.5 — judge preparation and transparent evaluation

Choose a tag to compare

@gowtham66867 gowtham66867 released this 27 Oct 18:26

FollowThrough v0.5 adds an invitation-gated shared fictional judge mode, a persistent inference-attempt ceiling and a source-precision guard for model-proposed deadlines. Judge hosting is not deployed. The public website remains a free fixture walkthrough.

Includes a five-minute judge guide, two transparent synthetic development evaluations (including a rejected prompt experiment), the remaining failure cases and a permissioned customer-pilot protocol. No customer pilot or production accuracy is claimed. Nineteen standalone checks and the frontend build passed.

The narrated 2:24 UI-capture montage is attached separately. The original YouTube upload still depicts v0.2; uploading the new video is blocked by the current browser account's YouTube eligibility.

Never include personal API keys in source or public demo assets. The test build's live mode requires configured credentials; free hosted judge access remains unresolved.

FollowThrough v0.4.0 — proposal impact map and final review

Choose a tag to compare

@gowtham66867 gowtham66867 released this 24 Oct 16:26

FollowThrough v0.4.0 — proposal impact map and actionable final review

Download followthrough-v0.4.0.zip, extract, create a Python 3.11+ virtual environment, install requirements.txt and run run.py. Open http://127.0.0.1:8787. The frontend is already built.

New:

  • Before you submit: one action queue for source review, missing or unreviewed answers, unsupported claims, missing attachments, deadlines and unresolved conflict decisions. Advisory owner assignments and exceptions are separate from blockers.
  • Impact map: trace source → requirement → selected evidence → answer → attachment/owner → package approval. Compare original and amended wording; filter changed obligations; open the exact source, evidence record, answer or owner form.
  • One backend readiness contract drives the screen, approval, export and user-reported submission. Expired evidence blocks an earlier approval even if its hash has not changed. Read-only projections do not alter revisions.
  • Missing-attachment actions preselect the deliverable; owner actions open its assignment form. Unanalyzed amendments explicitly show incomplete mapping.

Retained from v0.3: Qwen semantic evidence search followed by NVIDIA chat relevance assessment, human-reviewed NVIDIA vision transcription, and optional source-cited NVIDIA reasoning for conflicts. The dedicated Qwen reranker was unavailable on the tested account. Live features require your own Nebius key and YAML configuration. The explicitly fictional guided walkthrough works without credentials; no private developer key is distributed.

Validation: 13 standalone automated tests plus two development-only starter tests. Added contracts cover amendment/source/evidence/owner/approval links, restart persistence, expired evidence and consistent approval/export/submission gates. The browser walkthrough covered approval, a late amendment, changed wording and attachment targeting at desktop and narrow widths. Earlier v0.3 live smoke results are retained; no new inference path was introduced in v0.4.

Original public 2:30 demo (v0.2 core workflow): https://www.youtube.com/watch?v=TGrqRwF3qm0
The video does not show v0.3/v0.4 additions. See submission/12 through 17 screenshots for new features.

Single-owner local application, not public hosting or a multi-tenant service. Links reflect stored associations and model-proposed requirement keys; a person verifies authority, relevance and completeness. No agency pilots, measured savings, production accuracy, legal guarantee or autonomous bid submission are claimed. Optional Gmail OAuth remains unverified.

FollowThrough v0.3.0 — semantic evidence and visual conflict review

Choose a tag to compare

@gowtham66867 gowtham66867 released this 24 Oct 16:01

FollowThrough v0.3.0 — semantic evidence, visual sources and cited conflict review

Download followthrough-v0.3.0.zip, extract, create a Python 3.11+ virtual environment, install requirements.txt and run run.py. Open http://127.0.0.1:8787. The frontend is already built in the ZIP.

Load guided demo for a clearly labelled fictional walkthrough without accounts or credits. Live features require your own Nebius Token Factory key and YAML model configuration. No developer key or private user data is distributed.

New in this release:

  • Qwen embeddings shortlist approved, unexpired evidence; NVIDIA Nano reviews relevance. Candidates remain suggestions until selected and reviewed. The documented dedicated Qwen reranker returned model-not-found, so reranking explicitly uses NVIDIA chat relevance assessment.
  • NVIDIA Super VL transcribes scanned PDF pages, PNG/JPEG and document tables. Compare the transcription against each original page and confirm or correct it before analysis. At most six pages per visual import. This is model transcription with human review, not guaranteed OCR accuracy.
  • On-demand NVIDIA Super reasoning produces source-cited conflict and amendment findings. Findings require owner decisions; stale reviews block approval/export. No raw chain-of-thought is displayed or stored.

Core features retained: sourced requirements, PDF/DOCX/text imports, approved company evidence, missing-input tasks, amendment review, approval invalidation, DOCX/attachment export, separate submission and receipt evidence.

Validation: 11 standalone automated tests, including real HTTP/SQLite contracts, plus three live synthetic semantic/vision/reasoning smoke cases and six earlier extraction cases. These are not a production accuracy benchmark or customer impact measurement. See submission/advanced-live-evaluation.json and submission/live-evaluation.json.

Original core-workflow YouTube demo (v0.2): https://www.youtube.com/watch?v=TGrqRwF3qm0
This video does not show the v0.3 additions. New feature screenshots are included in submission/.

Single-owner local application. No public multi-tenant hosting, legal guarantee, automated external submission, arbitrary spreadsheet ingestion or independently authenticated receipts. Semantic evidence and visual transcriptions require human review. Gmail OAuth is optional and unverified. Do not expose this local app through a public tunnel.

See README.md for setup and model settings, submission/SUBMISSION.txt for the proposed Devpost entry, and docs/ARCHITECTURE.md for boundaries.

FollowThrough v0.2.0 — proposal readiness test build

Choose a tag to compare

@gowtham66867 gowtham66867 released this 23 Oct 07:00

FollowThrough v0.2.0 — local proposal-readiness test build

For judges: download followthrough-v0.2.0.zip, extract it, create a Python 3.11+ virtual environment, install requirements.txt, and run run.py. Open http://127.0.0.1:8787 and select Load guided demo. The frontend is already built in the ZIP.

No login or API key is needed for the explicitly labelled fictional guided walkthrough. Live NVIDIA analysis requires your own Nebius Token Factory key and config.yaml; full instructions are in README.md. No developer key or private user data is distributed.

Features: sourced requirements, PDF/DOCX/text import, approved company evidence, missing-input tasks, amendment impact review, approval invalidation, DOCX/attachment export, separate submission and receipt evidence.

Validation: 9 standalone tests passed; 11 including development-only starter coverage. Six small fictional live NVIDIA smoke examples passed strict schema, citation and expected-keyword checks. No production accuracy or customer impact is claimed.

Public YouTube demo: https://www.youtube.com/watch?v=TGrqRwF3qm0

The 2:30 narrated walkthrough is also attached as followthrough-demo.mp4. It uses actual application captures and explicitly fictional client/company information.

Boundaries: single-owner local application, no OCR or public multi-tenant hosting. Evidence excerpts require human relevance review. Gmail OAuth is optional and unverified. External submission remains manual. Do not expose this app publicly through a tunnel.

See submission/SUBMISSION.txt for the proposed Devpost entry and docs/ARCHITECTURE.md for limits.