Python Exam Figure Generator Service

We publish online mock tests for government exams (SSC, Railway) in English and Hindi. Today our team draws every question figure by hand. We need a Python service that draws them automatically from JSON, so our AI question generator can request a figure for each new question. The full specification is in the attached PDF (Project Brief). Please read it before bidding. BUILT WITH CLAUDE Designed to be completed by one developer using Claude (Claude Code) and free open-source Python libraries only: no paid software, no hand drawing, no external services. The sample zip includes CLAUDE.md (instructions for Claude Code) and PROJECT_SPEC.md (full spec in Markdown). All acceptance checks are automated. WHAT YOU WILL BUILD - A Python 3.10+ renderer: JSON in, clean print-quality SVG and PNG out, in our house style. - The JSON describes meaning, not coordinates ("tangent from P touching at R and S", "all A are B", "triangle with angles 30, 60, 90"). Your code works out every position. This needs real geometry: tangents, inscribed circles, triangles from sides and angles, 3D solids drawn in 2D. - A label placer: no label may ever touch a line, an arc or another label, even in crowded figures. This is checked by an automated test. - English and Hindi (Devanagari) labels, with correct conjunct shaping in SVG and PNG. - A JSON Schema per figure type, with separate errors for invalid input, an impossible figure and an unknown figure type. - Deterministic output: the same input always gives byte-identical files. - A REST API with an API key, packaged with Docker, plus a step-by-step deployment guide. - A JSON input for each of our 525 reference figures, plus a side-by-side review page (our figure next to yours). FIGURE TYPES (525 figures) Reasoning: Direction Sense (60), Non-Verbal Venn (53), Finding Missing Elements (44), Logical Venn (41), Coding-Decoding (37), Seating Arrangement (30), Series (17), Syllogism (14), Classification (3) Quant: Circle (65), Area (38), Volume (34), Triangles (26), Data Interpretation charts (12), Quadrilateral (16), Height & Distance (12), Lines & Angles (11), Coordinate Geometry (10), Mixture (2) 34 sample figures are attached. The full set (with manifest and errata) is shared after award. MILESTONES (quote a price and duration for each) M1 Core system and pilot (134 figures): core engine, label placer, schema framework, API, Docker, review page, test suite + Logical Venn, Non-Verbal Venn, Syllogism, Triangles M2 Reasoning and charts (205 figures): Seating Arrangement, Classification, Direction Sense, Coding-Decoding, Series, Finding Missing Elements, DI charts, Mixture M3 Geometry and handover (186 figures): Lines & Angles, Coordinate Geometry, Circle, Quadrilateral, Height & Distance, Area, Volume, final Docker package and deployment guide M1 is also the pilot. We release M2 and M3 to the same freelancer only if M1 is accepted. ACCEPTANCE Every figure in the milestone approved on the review page, the delivered zip builds with Docker and all tests pass (including the label-clearance and determinism checks), schemas and docs complete. Code is delivered as a zip through Freelancer. Final milestone: the API starts by following your deployment guide. All code and rights transfer to us on payment. No GPL or AGPL dependencies. HOW TO BID Start your proposal with the word FIGURE. Others will not be considered. Please include: 1. Links or screenshots of diagram, geometry or chart rendering you have built in code. 2. How you will keep labels clear of lines (the algorithm). 3. How you will render Devanagari correctly in SVG and PNG. 4. Your Python libraries, and how you will make the output byte-identical. 5. Your price and duration for each milestone, M1 to M3. 6. Confirmation that you accept the pilot-first arrangement. 7. Confirmation that you will build it with Claude Code and free open-source libraries only.