Technical Reference

Back to all tools

Every tool in the workspace is listed below with what it does, which AI model powers it, where your data is processed, and how each tool’s AI behaviour is guided.

Last updated: 2026-04-09 20:12 UTC

Table of Contents

Generate LO Documents

Category: lo Type: Step-by-step workflow URL: /brand

Description

Create branded reports, letterheads, and slide decks in LogicalOutcomes format. Upload content or paste text, download a polished file.

Model & Hosting

  • Model: google/gemini-2.5-flash-preview
  • Hosting: On-server (Canadian VPS) + OpenRouter API (metadata extraction for reports/letterheads, slide structuring)

Data Handling

  • For reports and letterheads, AI reads your document to identify the title, author, date, and structure. The formatting itself is done on our server without AI.
  • For slide decks, AI analyzes your full document and restructures the content into a presentation — choosing what goes on each slide, creating section breaks, and rewriting text to fit slide format.
  • AI processing uses Google Gemini via OpenRouter — your document content is sent to this service
  • All files are automatically deleted from our server after 1 hour

How It Works

AI reads the first portion of uploaded documents to identify titles, authors, dates, and other metadata. For slide decks, AI organises your text into structured slides with appropriate layouts (title slides, content slides, section dividers, statistics, quotes, and tables). Document formatting uses on-server tools (python-docx, pandoc, python-pptx) — not AI.

Full system prompt available on request for authorized reviewers.

Technical Notes

Metadata extraction and slide structuring use OpenRouter API with Gemini 2.5 Flash Preview (google/gemini-2.5-flash-preview). For reports and letterheads, the first ~4000 characters are sent to the LLM to identify document title, client, date, and author. Document formatting is done on-server using python-docx, pandoc, and python-pptx. Falls back to regex extraction if the LLM is unavailable.

Last verified: 2026-04-03

Transcribe

Category: pipeline Type: Step-by-step workflow URL: /transcribe

Description

Upload an audio or video recording (up to 2 hours) and get a text transcript with speaker labels. Takes a few minutes.

Model & Hosting

  • Model: Gladia transcription with pyannote speaker diarization
  • Hosting: On-server (Canadian VPS) + Gladia API (France)

Data Handling

  • Your recording is sent to Gladia (France-hosted) for transcription and speaker detection
  • Zero Data Retention — audio is processed in memory and never stored by Gladia
  • All files are automatically deleted from our server after 1 hour

System Prompt

(No system prompt — direct file upload to Gladia transcription API with diarization enabled. Speaker detection uses pyannote audio analysis, not an LLM.)

Technical Notes

Audio conversion via ffmpeg on-server. Transcription and speaker diarization use Gladia API (gladia.io, France-hosted, SOC 2 Type 2, ZDR by default). Speaker labels are generated from audio analysis (pyannote), not LLM inference. Files auto-deleted after session expiry (1 hour). Supports recordings up to ~2 hours.

Last verified: 2026-04-03

Remove Personal Information

Category: pipeline Type: Step-by-step workflow URL: /clean

Description

Upload transcripts, survey responses, or other text files. Pattern scanning catches phone numbers, emails, and SINs, then AI catches names, initials, job titles, and combinations that could identify someone. You review every flag before downloading.

Model & Hosting

  • Model: zai-org/GLM-4.5 (optional)
  • Hosting: On-server (Canadian VPS) + optional Nebius AI Studio EU API for AI review

Data Handling

  • By default, personal information is found using built-in scanning — your data doesn’t leave our workspace
  • You can optionally turn on AI review for a more thorough check (your text is sent to a privacy-focused AI service)
  • All files are automatically deleted from our server after 1 hour

How It Works

Built-in scanning uses pattern-matching (regex) to find and flag common personal information like phone numbers, email addresses, social insurance numbers, and postal codes — this runs entirely on-server with no AI involved. Optional AI review adds a second pass using a privacy-focused AI model trained to catch identifiers that patterns miss: names, initials, job titles that narrow identification, location references, anecdotes about specific events, and combinations of details that could identify someone in a small organisation. The AI applies Canadian privacy rules (PIPEDA) for de-identification and handles bilingual (English/French) content. All replacements use consistent markers ([NAME-1], [LOCATION-1], etc.) and are presented for your review before being applied.

Full system prompt available on request for authorized reviewers.

Technical Notes

Pattern-based PII detection uses regex only (no external calls). Optional AI review uses GLM-4.5 via Nebius AI Studio EU endpoint (direct API, no OpenRouter middleman) for privacy-focused processing.

Last verified: 2026-04-03

Qualitative Analysis

Category: pipeline Type: Step-by-step workflow URL: /qual

Description

Upload interview transcripts, survey responses, or CSV data. AI identifies patterns and suggests themes — you review and approve at every step. Exports a report with supporting quotes and a quality assessment.

Model & Hosting

  • Model: google/gemini-3.1-pro-preview
  • Hosting: On-server (Canadian VPS) + OpenRouter API

Data Handling

  • Your text is sent to Google Gemini (an AI service) to help find themes and draft findings. By this stage, you should have already used the Transcribe and Data Cleaning tools to remove names and other personal information from your files.
  • You review and approve every step before you download your Excel workbook.
  • All project data is automatically deleted from our server after 7 days — save your exports before then.
  • You can delete your project at any time from the Results page.

How It Works

The qualitative analysis workflow uses AI at several steps, with human review at every stage:

  • Auto-coding: AI reads your transcripts and identifies themes using thematic analysis (Braun & Clarke). If you provide research questions, it uses them as deductive coding categories while also looking for emergent themes. All excerpts are verbatim quotes attributed to specific participants.
  • Theme suggestions: After initial coding, AI looks for patterns not yet captured — uncoded passages, cross-transcript connections, minority voices, and disconfirming evidence.
  • Theme refinement: AI generates descriptive theme names (using participant language), narrative descriptions, within-theme variation, tensions, and silences.
  • Synthesis: AI produces a thematic map showing how themes relate, an overall narrative, and documentation of what the data does not say.
  • Quality review: AI assesses each theme’s quality — name quality, variation, disconfirming evidence, demographic claims, and flagged phrases.

All AI outputs are presented for your review before being applied. Source locations for excerpts are determined by fuzzy matching on-server, not by the AI model.

Full system prompt available on request for authorized reviewers.

Technical Notes

All coding, theme generation, refinement, synthesis, and quality review use OpenRouter API with Gemini 3.1 Pro Preview (google/gemini-3.1-pro-preview). All AI outputs are presented for human review before being applied. Source locations for excerpts are determined by post-hoc fuzzy matching, not LLM-reported positions.

Last verified: 2026-04-03

LO Guide Assistant

Category: lo Type: Step-by-step workflow URL: /handbook

Description

Search LogicalOutcomes policies, procedures, and guides. Ask questions in plain language and get answers with links to the source.

Model & Hosting

  • Model: gemini-2.5-flash
  • Hosting: Google Gemini API (File Search)

Data Handling

  • Your questions are sent to Google’s Gemini AI to search the LO Guide
  • Answers are grounded in handbook content only — the AI does not use external knowledge for policy claims
  • No conversation data is stored permanently — sessions expire after 1 hour
  • Avoid including personal or sensitive details in your questions

How It Works

AI searches the LogicalOutcomes handbook using Gemini File Search (RAG). Answers are grounded in retrieved handbook content. The system prompt instructs the model to translate technical language into plain language, cite sources, and acknowledge when coverage is incomplete.

Full system prompt available on request for authorized reviewers.

Technical Notes

Uses Google Gemini File Search for retrieval-augmented generation over ~300 markdown files from the LO Guide wiki. Citations link to guide.logicaloutcomes.net/en pages.

Last verified: 2026-04-05

Learn the Platform

Category: more-tools Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=ai-guide

Description

Start here if you’re new. Asks about your work, then walks you through five parts — from safety basics to writing your first draft to using expert panels. Adapts to your pace and uses your real work as practice material.

Model & Hosting

  • Model: google/gemini-3-flash-preview
  • Hosting: Open WebUI (self-hosted) + Google Gemini API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

The AI Guide walks new users through a structured first lesson covering:

  • Safety foundations — what’s safe to share with AI (traffic light guide: red/yellow/green) and the principle that all AI output is a draft
  • Interface orientation — navigating the platform, choosing models, and finding features
  • First task — a quick real-world draft to build confidence, with a four-point quality check
  • Improvement loop — comparing quick drafts with careful, considered responses
  • Expert panel — demonstrating the Convening Experts skill for multi-perspective analysis
  • Verification — fact-checking claims and using web search appropriately

The Guide adapts to each learner’s pace and role, uses nonprofit examples throughout, and teaches the CALM checklist (Confidential? Ask clearly. Look for assumptions. Make it true). It also includes a document review mode for practising with sample nonprofit documents.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search enabled. Max 4096 tokens. Interactive lesson covering safety, interface, drafting, improvement, and verification.

Last verified: 2026-04-03

LO Evaluation Framework Builder

Category: lo Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=evaluation-framework

Description

Build a complete evaluation framework interactively: logic model, evaluation questions, indicators, data collection plan. CES-aligned, uses LO methodology. Revise section by section before generating the final document.

Model & Hosting

  • Model: google/gemini-3.1-pro-preview
  • Hosting: Open WebUI (self-hosted) + Google Gemini API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

The Evaluation Framework Builder follows the LogicalOutcomes evaluation framework (Kerr & Llewelyn, 2024), which treats evaluation as an intervention that improves services — not just a reporting exercise.

When you describe your program, the builder:

  • Researches your organisation (via web search, if enabled) to understand your programs, mission, and the people you serve
  • Generates a complete evaluation framework with six sections: program overview, logic model, evaluation questions, interest groups analysis, indicators, and next steps
  • Uses the organisation’s own language for the people they serve
  • Prioritises participant tracker data and small-sample interviews over surveys
  • Ensures every recommended measure connects to a specific program improvement action
  • Builds capacity incrementally (outputs before outcomes, short-term before long-term)

The framework is generated in Markdown for collaborative review, then converted to a formatted HTML document. Based on the LogicalOutcomes Evaluation Planning Handbook by Gillian Kerr and Sophie Llewelyn.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search enabled. Max 8192 tokens. Follows LogicalOutcomes evaluation framework (Kerr & Llewelyn, 2024). Can research organisations via web search.

Last verified: 2026-04-03

Analyze Documents

Category: more-tools Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=document-analyst

Description

Upload reports, research papers, strategic plans, or data files and ask questions. Summarize findings, find patterns across documents, identify gaps, or compare what different sources say. Handles multiple documents in one conversation.

Model & Hosting

  • Model: z-ai/glm-5
  • Hosting: Open WebUI (self-hosted) + ZHIPU AI GLM-5 API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

The Document Analyst reads, analyses, and synthesises uploaded documents — evaluation reports, program documents, research papers, interview transcripts, data sets, strategic plans, and clinical outcomes reports.

Capabilities include:

  • Document summary — key findings, methodology, conclusions, and limitations
  • Pattern and theme identification — recurring themes with supporting evidence and counter-evidence
  • Gap analysis — missing topics, unanswered questions, contradictions between sections
  • Cross-document synthesis — agreement, disagreement, and complementary findings across multiple documents
  • Data interpretation — trends, anomalies, and whether data supports narrative claims
  • Qualitative analysis support — first-pass coding, consistency checks, evidence assembly, and counter-evidence surfacing (always marked as drafts for analyst review)

Privacy-focused by default: uses participant codes (P1, P2) instead of real names in outputs, flags identifying information, and reminds users about de-identification. Clearly labels what a document states vs. implies vs. what the model infers. Web search is disabled to prevent mixing external content into document analysis.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search disabled (prevents mixing external content into document analysis). Max 8192 tokens. Privacy-focused: flags identifying information, uses participant codes.

Last verified: 2026-04-03

Quick Assistant

Category: more-tools Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=quick-and-superficial

Description

Fast first pass on anything — summaries, draft content, concise options with tradeoffs. Makes reasonable assumptions and moves quickly. Flags when something needs deeper analysis.

Model & Hosting

  • Model: google/gemini-3-flash-preview
  • Hosting: Open WebUI (self-hosted) + Google Gemini API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

Quick Assistant provides fast, lightweight, first-pass outputs — quick summaries (3-7 bullets), short structured drafts, and concise options with tradeoffs. Makes reasonable assumptions when context is missing and states those assumptions. Flags when a request needs deeper analysis and suggests the appropriate specialist model. Uses Canadian spelling and metric measurements.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search enabled. Max 2048 tokens. Lightweight, fast outputs. Makes reasonable assumptions when context is missing.

Last verified: 2026-04-03

Thoughtful Assistant

Category: more-tools Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=thoughtful-and-slow

Description

Asks clarifying questions first, then thinks from multiple angles. Separates what’s known from what’s inferred, challenges assumptions, catches US-default advice and flags Canadian equivalents. For complex decisions where getting it right matters more than speed.

Model & Hosting

  • Model: z-ai/glm-5
  • Hosting: Open WebUI (self-hosted) + ZHIPU AI GLM-5 API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

Thoughtful Assistant takes a careful, multi-angle approach to complex questions:

  • Asks before answering — 2-3 clarifying questions to understand what you actually need, what you already know, and what constraints exist
  • Considers multiple perspectives — what the evidence says, what experience suggests, who might disagree, what’s missing, and what could go wrong
  • Separates knowledge from inference — clearly labels claims as established knowledge, reasonable inference, or speculation
  • Challenges assumptions gently — names assumptions that might be wrong, like a colleague pointing something out
  • Flags Canadian context — catches U.S.-default assumptions (PIPEDA not HIPAA, AODA not ADA, CRA not IRS) and notes the Canadian equivalent

Designed for questions where depth and nuance matter more than speed. Uses evidence-based language throughout.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search enabled (deepr_web_research tool). Max 8192 tokens. Asks 2–3 clarifying questions before responding. Canadian context aware (PIPEDA, AODA, CRA).

Last verified: 2026-04-03

Evaluation Planner

Category: more-tools Type: Step-by-step workflow URL: https://evalplanner.logicaloutcomes.net/

Description

Describe your program and provide your website. AI researches your organization, asks clarifying questions, and produces a complete evaluation plan with logic model, evaluation questions, and indicators. Download as HTML or receive by email.

Model & Hosting

  • Model: OpenRouter (configurable per step)
  • Hosting: Self-hosted (Canadian VPS) — evalplanner.logicaloutcomes.net

Data Handling

  • Program details are processed by AI to generate your evaluation plan (data passes through OpenRouter, a US-based AI routing service)
  • Your website may be read automatically to gather program context
  • Reports can be emailed to you or downloaded as HTML

How It Works

Enter your program details and website. AI researches your organisation, asks clarifying questions, and generates a complete evaluation plan with logic model, evaluation questions, indicators, and next steps. Reports can be emailed or downloaded as HTML.

Full system prompt available on request for authorized reviewers.

Technical Notes

React/TypeScript SPA with Express backend. AI prompts are managed in a PostgreSQL database. Background job queue handles long-running AI calls. Reports rendered as responsive HTML with logic model diagrams.

Last verified: 2026-04-03

LO Contract & Proposal Drafter

Category: lo Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=contract-drafter

Description

Draft complete proposals and contracts using LO templates. Fixed-price, time-and-materials, or custom. Produces ready-to-send documents, not templates with blanks.

Model & Hosting

  • Model: anthropic/claude-sonnet-4.6
  • Hosting: Open WebUI (self-hosted) + Anthropic Claude API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

Drafts complete proposals and contracts from LogicalOutcomes templates. Describe a potential engagement and the model produces a full proposal or contract, flags items needing your input, and can generate a branded LO Word document. Also walks through contract review when you receive a contract from a client.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search disabled. Has loaded templates: proposal template, fixed-price contract, time-and-materials agreement, contract review procedure. Branded document tool (lo_brand_document) available.

Last verified: 2026-04-03

LO Project Starter

Category: lo Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=project-starter

Description

Generate project charters, role assignments, and kick-off meeting agendas from LO templates. Complete documents in one conversation.

Model & Hosting

  • Model: anthropic/claude-sonnet-4.6
  • Hosting: Open WebUI (self-hosted) + Anthropic Claude API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

Walks consultants through the LO project setup sequence: charter, roles, and kick-off. Paste a proposal or project description and the model drafts a complete project charter, assigns roles, and plans the kick-off meeting using LO templates and examples.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search disabled. Has loaded templates: project charter, two charter examples, kick-off agenda, project roles, role terms. Branded document tool (lo_brand_document) available.

Last verified: 2026-04-03

LO Consulting Assistant

Category: lo Type: AI assistant (Open WebUI) URL: https://ai-dev.logicaloutcomes.net/?model=lo-consulting

Description

Your daily working assistant. Evaluation reports, logic models, research, strategic planning, facilitation design. LO methodology, CES standards, and procedures built in.

Model & Hosting

  • Model: anthropic/claude-sonnet-4.6
  • Hosting: Open WebUI (self-hosted) + Anthropic Claude API

Data Handling

  • Conversations saved in your AI Chat account
  • You can delete conversations at any time

How It Works

General-purpose daily assistant for LO consultants. Supports evaluation design, report writing, data analysis, strategic planning, facilitation design, and stakeholder engagement. Has LO’s evaluation methodology, consent forms, research ethics policy, and procedures loaded. Produces complete drafts (not templates), connects to specialist skills for structured workflows, and can generate branded LO Word documents.

Full system prompt available on request for authorized reviewers.

Technical Notes

Web search enabled. Max 8192 tokens. Has loaded documents: consent forms, research ethics policy, evaluation frameworks, privacy policies. Branded document tool (lo_brand_document) available.

Last verified: 2026-04-03

LO Guide

Category: lo Type: AI assistant (Open WebUI) URL: /handbook

Description

Quick lookup of LO policies, procedures, templates, and organizational information. Has the full document library loaded.

Model & Hosting

  • Model: gemini-2.5-flash
  • Hosting: LO Workflow Tools (Canadian VPS) + Google Gemini API

Data Handling

  • Your questions are sent to Google Gemini for answering, with relevant handbook sections retrieved automatically
  • The LO document library is pre-indexed in Gemini File Search — only relevant excerpts are used per query

How It Works

Answers questions about LogicalOutcomes policies, procedures, templates, and organizational information using the full LO document library as context.

Full system prompt available on request for authorized reviewers.

Technical Notes

Built-in RAG assistant using Gemini 2.5 Flash with File Search. Handbook indexed via CLI indexer. Citations map to handbook wiki URLs.

Last verified: 2026-04-05