Colectica-routing-toolkit

It can also ingest a Forsta+ (Confirmit Horizons) XML export of the same questionnaire wave and structurally compare its routing against the Colectica-derived routing, surfacing any discrepancies — missing branches, unmatched conditions, Forsta+-only "else" branches — in a dedicated routing-diff GUI, with a side-by-side graph view per discrepancy.

Django owns all data, routing logic, and validation. Flowise is advisory only.

Flowise (an external LLM agent platform) is used for two narrow purposes, and its output is always validated/post-processed by Django before being trusted:
- Module Review agentflow— reviews routing/coverage for design issues. Django sends a compact payload and post-validates the response against known facts, rejecting anything that invents question names or modifies routing.
- Interview Wording agentflow— reformats respondent-facing question text/options during the interview simulator. Django validates the response and falls back to a deterministic, locally-built message if Flowise is unavailable or returns something invalid — the simulator always works even if Flowise is down.

Enforced in the UI/views as a strict order per uploaded module:
- UploadJSON →QuestionnaireModule
- Extract schema(schema_extractor.ColecticaSchemaExtractor) →NormalizedQuestionrows
- Extract routing(routing_extractor.ColecticaRoutingExtractor) →RoutingEdgerows (conditional / sequential / loop)
- Build graph(graph_builder.py+graph_enrichment.py) →QuestionnaireGraph(nodes/edges JSON + Mermaid text)
- Run Flowise review(optional) →ModuleAIReview

Everything runs synchronously in the request/response cycle — there is no Celery/async task queue.

A second, independent pipeline runs against a secondQuestionnaireModule(source_formatauto-detected asforsta_xmlfrom the.xmlextension at upload — same upload form as Colectica), cross-validating a fieldwork agency's Forsta+ (Confirmit Horizons) XML export against the Colectica-derived routing for the same wave:
- UploadForsta+ XML →QuestionnaireModule
- Extract schema(forsta_xml_schema_extractor.ForstaXmlSchemaExtractor) →NormalizedQuestionrows
- Extract routing(forsta_xml_routing_extractor.ForstaXmlRoutingExtractor) →RoutingEdgerows
- Match questions(question_matcher.build_question_matches) →QuestionMatchrows, pairing each Colectica question with its best Forsta+ counterpart in three passes: exact normalized-text match, then fuzzy fallback (difflib, 0.75 threshold), then aname-tiebreakreconciliation step — if a question's current match has a different name than itself, and an unused same-named question exists on the other side whose own wordingalsoclears the fuzzy threshold (or is a clean prefix/substring match — Forsta+ source text sometimes folds interviewer instructions inline where Colectica keeps them separate), the same-named one takes over. Name is a tiebreak, never an override: a same-named-but-unrelated-content "false friend" is left alone.
- Compare routing(routing_comparator.compare_routing_for_modules) →RoutingDiscrepancyrows — astructuraldiff (edge target presence only, not condition semantics). Both the source and target question of each edge are resolved throughQuestionMatchbefore comparing, not compared as raw name strings, so a target present on both sides under a different name (casing, a Forsta+ suffix, etc.) isn't wrongly reported as missing.

Browsable at/questionnaires/routing-diff/<colectica_module_id>/<forsta_module_id>/, with a per-discrepancy detail page rendering both systems' routing graphs side by side.

python manage.py runmcp # own process, port 8765 by default

Every request must carry a per-person access token in its URL path (https://<host>/t/<token>/mcp) rather than a header, since MCP client connector UIs generally only take a URL. Staff users generate/revoke tokens at/questionnaires/mcp-tokens/— no shared secret, no terminal command needed to onboard a new person, and revoking one person's token doesn't affect anyone else's.

Requesting access:there's no self-serve signup by design. Open an issue on this repository or contact the maintainer to request a token.

Tools (mcp_server/tools.py), each with a matching MCPprompt(mcp_server/prompts.py) that MCP clients can surface as a slash-command-style shortcut:

Never writes to the database and never triggers a compute-heavy pipeline step (extraction, graph building, matching/comparison, AI review) itself — every tool reads data some other part of the app already computed and persisted.

- Python 3.12+
- PostgreSQL, with aflowise_questionnaire_dbdatabase available
- A running Flowise instance (optional — only needed for the AI review / interview wording features; the rest of the app works without it)

git clone https://github.com/amiravarzamani/colectica-forsta-routing-toolkit.git cd flowise-questionnaire-system python3 -m venv venv source venv/bin/activate # venv\Scripts\activate on Windows pip install -r requirements.txt

Copy.env.exampleto.envand fill inSECRET_KEY/DB_PASSWORD/DB_HOST—config/settings.pyhas no defaults for these and will fail loudly at startup if they're missing. Adjust theDATABASESandFLOWISE_settings inconfig/settings.pyto match your environment before running migrations.

python manage.py migrate python manage.py createsuperuser # first user, since login is required app-wide python manage.py runserver

config/ Django project settings, URLs, WSGI/ASGI flowise_questionnaire/ models.py QuestionnaireModule, NormalizedQuestion, RoutingEdge, QuestionnaireGraph, ModuleAIReview, InterviewSimulatorSession/Turn, QuestionMatch, RoutingDiscrepancy services/ Pipeline logic, in order: schema_extractor.py parse questions out of the Colectica JSON routing_extractor.py parse conditional/sequential/loop routing (Colectica) forsta_xml_schema_extractor.py parse questions out of the Forsta+ XML forsta_xml_routing_extractor.py parse conditional/sequential/loop routing (Forsta+) graph_builder.py build the routing graph graph_enrichment.py annotate the graph condition_evaluator.py evaluate Colectica-syntax routing conditions against answers forsta_condition_evaluator.py evaluate Forsta+-syntax routing conditions against answers coverage_intent_builder.py generate deterministic test-case seed inputs routing_simulator.py check routing coverage question_matcher.py pair Colectica and Forsta+ questions (exact + fuzzy + name-tiebreak) routing_comparator.py structural diff of matched questions' routing edges (source + target resolved via QuestionMatch) routing_diff_explainer.py plain-language explanation text for the routing-diff GUI agentflow_payload_builder.py build the Module Review Flowise payload flowise_client.py send/receive the Module Review agentflow interview_router.py deterministic routing engine for the simulator interview_simulator_service.py orchestrate simulator sessions answer_validation.py validate respondent A/B/C input question_presentation.py convert questions to respondent-facing text flowise_interview_wording.py Interview Wording agentflow client + caching/fallback interview_simulator_contracts.py shared dataclasses views/ module_views.py upload / extract / build-graph / review / graph interview_simulator_views.py start / state / answer / abandon routing_simulation_views.py routing_diff_views.py Colectica-vs-Forsta+ report / run / discrepancy-detail auth_views.py mcp_server/ models.py McpAccessToken (per-person access token) auth_middleware.py TokenAuthMiddleware -- validates /t/<token>/mcp on every request tools.py the MCP tools (see "MCP tool server" above) prompts.py matching MCP prompts (slash-command shortcuts) server.py MCPServer instance, tool/prompt registration views.py / urls.py staff-only token management UI (/questionnaires/mcp-tokens/) management/commands/runmcp.py standalone streamable-HTTP server command

No license file yet — all rights reserved by default until one is added.

Transaction-complete hotel booking over MCP — 300K+ properties, real hotel confirmation numbers, loyalty points, secure checkout. Hotels are merchant of record. Builders set their own booking fee via Stripe Connect. Built on proven distribution infrastructure.

An MCP server for AI video generation. MCP server for AI video generation. Lets Claude, ChatGPT, OpenClaw , Hermes & other agents create AI videos and publish them to YouTube, TikTok, Instagram etc..

Institutional research and manager diligence reports on hedge funds, venture capital and private equity managers. Summary of filings, personnel changes, media screening and social signals delivered to you in minutes.

ALTER - identity infrastructure for the AI economy

D2C eCommerce fulfillment platform: manage orders, inventory, shipments, campaigns, and billing via AI agents

Apigene MCP Gateway is the runtime layer that connects AI agents to APIs and MCP servers via Model Context Protocol.

MCP to interface with multiple blockchains, staking, DeFi, swap, bridging, wallet management, DCA, Limit Orders, Coin Lookup, Tracking and more.

MCP server for Bitnovo Pay integration with AI agents. Provides cryptocurrency payment capabilities through Bitnovo Pay API. Features include payment creation, status checking, QR code generation, and webhook management with support for multiple tunnel providers (ngrok, zrok, manual).

Shop for gift cards, esims, phone topups. Pay with cards and crypto.