MindMap Debugger is a prototype that extracts claims and relationships from pasted text, then flags contradictions and circular reasoning. In a retrospective about building it for the WeMakeDevs × AWS First Commit Build It track, Sagar Maurya describes a three-day lesson: combining several plausible model outputs can create errors that no individual run contained. His fixes focused not just on extraction, but on how claims and relations were merged and checked.
What MindMap Debugger does
The tool takes pasted arguments or transcripts and asks a language model to identify propositions and relations between them. The relations Maurya describes include supports, depends_on, and contradicts. The application checks contradiction links and searches for circular reasoning across depends_on and supports connections. It presents findings in a 3D relation graph and a plain-language summary.
As an Amazon Associate I earn from qualifying purchases.
Maurya describes a pipeline that runs extraction three times through Groq using the Strands Agents SDK, merges similar propositions, processes relations across the runs, deduplicates cycles, and applies a Cedar policy gate to findings. The stack he names is Strands Agents SDK, Cedar, Groq’s gpt-oss-120b, Flask, and Three.js. This is an account of the implementation, not an independent evaluation of how accurately it handles arbitrary text.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Day one: get text through the pipeline
Maurya’s first version connected pasted text to model-based claim extraction and relation detection. It could reportedly extract claims, identify an obvious contradiction, and display a rough interface. But the model initially returned internal reasoning text instead of clean JSON, so the output was not in the format the rest of the application needed.
After other parameter placements failed, Maurya says setting reasoning_format=hidden and reasoning_effort=high through Strands’ extra_body parameter resolved that output problem. He chose Groq because it was available to him without a payment card; AWS account signup and billing constraints prevented him from using Bedrock. Strands let him use Groq’s OpenAI-compatible endpoint, while Cedar served as a policy gate for displaying findings.
Day two: make repeated extractions useful
The same input did not always yield the same findings. Maurya reports that five repeated runs returned between two and six findings. He attributed the variation to non-deterministic serving of gpt-oss-120b and changed the pipeline to extract three times and merge the results. Repeating extraction gave him multiple views of the text, but it also made the merge rules consequential: differences between runs could be useful variation, or could become false claims and cycles if combined carelessly.
Matching claims without missing paraphrases
Exact string matching failed to merge paraphrases that expressed the same proposition. A similarity-based approach was needed to recognize that differently worded claims could be equivalent. But a loose similarity measure introduced its own risk, which became clear on day three.
Rank #2
- 【jigsaw puzzles 1000 pieces for adults 】Reading Library Books adult puzzles Includes 1000 piece puzzle for adults clearance and a bonus glare-free puzzle poster; coolest adult puzzles with exquisite packing box . jigsaw puzzles for adults with letters on back helps you complete the puzzles more effectively, adult puzzles 1000 pieces and kids ages 14+. Finished Size: 27.56" x 19.69"/70*50cm
- 【GREAT MATERIAL】hard Book library puzzles for adults is made of premium recycled cardboard , sturdy and durable, allowing for multiple uses. Each piece is precisely cut ensuring a perfect fit，and advanced technology ensuring Dust-free. fully interlocking,Each puzzle piece is unique shape providing a satisfying challenge for puzzle lovers , rounded edges comfortable to the touch, fit well and safe for everyone
- 【1000 piece puzzles】 puzzles 1000 piece Improve hand-eye coordination and develop problem solving skills, logical thinking and patience ability,provide hours of fun and entertainment for your family and friends ,a fun night,offering relaxation, stress reduction. A great way to connect with family members and friends to develop a closer relationship. Feel the sense of accomplishment and pride in completing this unique puzzles for adults
- 【PERFECT GIFT】High-quality difficult puzzles for adults puzzle is wrapped in a sturdy, gift-ready box,cool puzzles for adults Is the perfect gift for Birthday and Christmas thanksgiving gifts for kids, lover, friend, colleagues and parents. After the puzzle is completed, 1000 piece puzzles for adults can be used as a mural (Suggested frame size: length 70.3cm/width 50.3cm),to decorate your living room,the beautiful and coolest Home Decora funny puzzles for adults
- 【 PIECE MISSING SUPPORT】: If you have Piece Missing Issue about our puzzles 1000 pieces , just feel free to contact us, we’ll replace it with the same image, try our best to help you solve the issue. fun adult games for game night,family puzzles for kids and adults.
Looking for cycles across relation types
The first cycle search followed only depends_on edges, so it missed loops that mixed dependency and support relations. Maurya says he changed cycle detection to consider both depends_on and supports edges. Depth-first search could also find the same cycle from different starting nodes; he deduplicated those results by their node sets.
Day three: stop the merge from inventing findings
The most important failures in Maurya’s account came from combining individually plausible outputs. A merge could create a claim that was too broad, or a graph with a cycle that no single extraction run had proposed. His fixes addressed those two failure modes separately.
Use Jaccard similarity to avoid merging distinct claims
Maurya reports that a word-overlap score divided by the smaller set of words merged two different witness-testimony propositions because they shared common words. He replaced it with Jaccard similarity: the number of shared words divided by the size of the union of the two word sets. In his example, six shared words in a union of twelve produce a score of 0.5, below the reported 0.7 merge threshold.
Rank #3
- SIZE: The size of library puzzle is 23.85*23.85 inches/60.6*60.6 cm (after completed). The size of the packaging box is 7.87*7.87*2.36 inches. Sturdy packaging box to keep every puzzle piece intact. And a poster is attached for reference
- MAKE THE PUZZLE EASY: The back of nature puzzle is divided into several parts by letters. If you think the puzzle is too difficult, you can complete the puzzle according to the letter division
- YOU WILL LOVE THIS PUZZLE: Landscape puzzle is made of thickened cardboard, printed with colored ink. Precise fit, no puzzle dust. Sturdy puzzle pieces ensure multiple assembly without bending or deformation
- DIFFICULT PUZZLE: Book puzzles for adults allows you to enjoy the fun of putting together jigsaw puzzles. Once the puzzle is completed, you can proudly display it on your wall and let its bright colors bring charm to your space
- FAMLIY PUZZLE: This book theme jigsaw puzzles is for puzzle lover. It can not only be used as a decoration, but also you can play puzzles with family and friends during holidays or parties and spend a happy time
In a four-sentence test, Maurya says the change preserved four propositions, avoided a false self-loop, and retained the intended cycle involving three claims. This is an example from his retrospective, not a general accuracy benchmark; similarity scores can help control merging, but the example does not establish how the method performs across different writing or vocabulary.
Recommended Free Tools
Resolve reversed dependency edges before searching for cycles
Different extraction runs sometimes reversed the direction of a depends_on relation in Maurya’s bridge-maintenance example. Simply unioning the edges from all three runs put both directions into the combined graph, creating circular findings that were absent from any individual run. Maurya says he pruned reverse-direction dependency pairs, keeping the direction with higher confidence.
In that reported example, the result changed from one contradiction and four circular findings to one contradiction and no circular findings. The result illustrates why aggregating outputs is not automatically safer: if runs disagree about edge direction, the combined graph can assert a loop that none of them individually contained.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why a working pipeline and polished interface were not enough
A pipeline can complete successfully while producing incorrect findings. Likewise, a clean-looking graph does not prove that the claims, merged relations, or detected cycles are sound. Maurya’s account points to two practical checks: test deliberately chosen edge cases, and inspect raw logs rather than judging only by the final screen.
- Test paraphrases as well as exact duplicates to see whether equivalent claims merge without collapsing distinct ones.
- Include cycles that mix support and dependency relations, not only cycles made of one edge type.
- Check whether a reported cycle appears in an individual run or only after combining conflicting directions.
- Inspect extraction outputs and merge decisions alongside the rendered graph and summary.
These checks reflect the failures Maurya encountered; his retrospective does not provide an independent benchmark, a complete test suite, or evidence that the reported accuracy generalizes to other text.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →The interface bug: hide the canvas until it is ready
Not every bug involved model behavior. Maurya reports a white flash on initial page load because the canvas painted before WebGL had rendered a frame. His described fix kept the canvas transparent until the first rendered frame, avoiding the premature white display.
Best Value
What the retrospective establishes—and what it does not
Maurya’s account shows how this prototype evolved from a basic extraction flow into a system that tried to account for variation between model runs. It documents specific implementation choices and examples: three extractions, Jaccard-based proposition merging, cycle detection across support and dependency links, cycle deduplication, and pruning reversed dependency pairs by confidence.
It does not establish a general success rate, a reproducible evaluation across varied arguments, or the current status of a hosted product. The reported test outcomes should be read as author-described examples rather than independently verified product results. The article says the repository is MIT licensed and gives local run instructions: pip install flask strands-agents openai, then python app.py; those details are likewise statements in the retrospective, not independently checked repository information.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
