Evaluated on FinMME (Luo et al., 2025) — we do not own this dataset.
+
+
+
+
+
+
+ Figure 1. Agent recovers multi-select answers zero-shot misses. On a FinMME GNPA chart, zero-shot selects a single wrong period (1HFY23, 5.1%), while AgentFinVQA applies a Plan → OCR → Legend → Vision → Verify pipeline, audits each MCQ option against the 4% threshold, and recovers the full answer (1QFY24 + 9MFY24). The full reasoning trace is stored as a Model Evaluation Packet (MEP).
+
+
+
+
+
+
+ AgentFinVQA is a multi-agent framework for evaluating vision-language models on financial chart understanding.
+ It decomposes chart QA into an explicit Plan → Inspect → Verify loop,
+ producing fully traceable Model Evaluation Packets (MEPs) for reproducible,
+ post-hoc analysis across VLM backends.
+
+ Financial chart question answering in regulated settings demands more than accuracy: practitioners
+ must know which answers to trust before acting on them, and many institutions cannot send client
+ data to external model providers. Yet existing chart-QA agents are accuracy-focused and opaque,
+ and most assume proprietary API access; none is both auditable and deployable on-premise.
+
+
+ We present AgentFinVQA, a multi-agent pipeline that decomposes each query into
+ planning, OCR, legend grounding, visual inspection, and verification, recording every step in a
+ traceable Model Evaluation Packet (MEP) per sample. On FinMME, AgentFinVQA
+ improves +7.68 pp over a model-matched zero-shot baseline with a proprietary
+ backbone (Gemini-3; 71.24% vs. 63.56%, McNemar p ≈ 1 × 10⁻¹⁶), and
+ +4.84 pp with open-weights Qwen3.6-27B-FP8 served locally on a single A100
+ at roughly one-tenth the cost, so the gains do not depend on a proprietary API.
+
+
+ The verifier's verdict also serves as a useful confidence signal (68.2% vs. 55.6% exact
+ accuracy on confirmed vs. revised answers), enabling human-in-the-loop review routing.
+ Together these results show that auditable, on-premise financial chart QA is practical and that
+ the open-weights system keeps most of the accuracy gains at a fraction of the cost. We release
+ our code and per-sample MEPs to support reproducible evaluation.
+
+
+
+
+
+
Method
+
+ AgentFinVQA coordinates four specialized components in a structured pipeline.
+ Unlike single-pass VLM approaches, each component has a clearly scoped role,
+ and the entire execution trace is captured for post-hoc explainability.
+
+
+
+
+
1
+
+
PlannerAgent
+
Text-only LLM that generates a structured JSON inspection plan — without seeing the image.
+ Separates strategic reasoning from visual perception.
+
+
+
↓
+
+
2
+
+
OcrReaderTool
+
Focused VLM call that transcribes all visible chart text (axes, labels, legend, annotations)
+ into structured JSON, grounding downstream agents in observed text rather than hallucinated values.
+
+
+
↓
+
+
3
+
+
VisionAgent
+
CrewAI-orchestrated agent that executes the planner's steps using vision_qa_tool
+ to produce a draft answer and explanation.
+
+
+
↓
+
+
4
+
+
VerifierAgent
+
Second VLM that critically reviews the draft answer against the chart image and issues a
+ CONFIRM or REVISE verdict with confidence score.
+
+
+
↓
+
+
✓
+
+
Model Evaluation Packet (MEP)
+
Portable JSON artifact capturing the full trace: inspection plan, OCR output, vision reasoning,
+ verifier critique, tool call logs, timestamps, and per-stage errors.
+ Enables reproducible evaluation and model comparison across backends.
+
+
+
+
+
+
Supported Backends
+
All agents support swappable VLM backends via a unified interface:
+
+
Gemini — Google Gemini 3 Flash / 2.5 Flash (default)
+
OpenAI-compatible — GPT-4o, or any locally-served model via vLLM (e.g. Qwen3.5 / Qwen3.6)
+
Planner and Verifier can use different backends from the VisionAgent
+
+
+
+
+
+
Dataset
+
+
+
FinMME
+
+ AgentFinVQA is evaluated on FinMME, a challenging financial multimodal benchmark
+ spanning 18 financial domains and 6 asset classes. Samples are sourced from real-world financial
+ research reports and cover three cognitive levels, comprehensive understanding, fine-grained
+ perception, and analysis & reasoning, validated by 20 expert annotators.
+
+
+
10 major chart types, 21 subtypes (bar, line, pie, candlestick, and more)
+ AgentFinVQA consistently outperforms same-model zero-shot baselines across both evaluated backbones on FinMME.
+ With Gemini-3 Flash the agent gains +7.68 pp mean accuracy
+ (p ≈ 1.1 × 10⁻¹⁶); with the locally-served
+ Qwen3.6-27B-FP8 (vLLM) it gains +4.84 pp
+ (p ≈ 3.0 × 10⁻⁶), demonstrating backend-agnostic improvements.
+
+
+
+
+
+
+
System
+
Backbone
+
Mean Acc. ↑
+
Exact Acc. ↑
+
MCQ Mean ↑
+
Std Mean ↑
+
+
+
+
+
Zero-shot
+
Gemini-3 Flash Preview
+
63.56%
+
56.40%
+
64.4%
+
54.0%
+
+
+
AgentFinVQA
+
Gemini-3 Flash Preview
+
71.24%
+
65.12%
+
72.5%
+
57.0%
+
+
+
Δ
+
+
+7.68 pp
+
+8.72 pp
+
+8.1 pp
+
+3.0 pp
+
+
+
Zero-shot
+
Qwen3.6-27B-FP8
+
61.68%
+
53.52%
+
62.8%
+
49.0%
+
+
+
AgentFinVQA
+
Qwen3.6-27B-FP8
+
66.52%
+
60.24%
+
68.1%
+
48.0%
+
+
+
Δ
+
+
+4.84 pp
+
+6.72 pp
+
+5.3 pp
+
−1.0 pp†
+
+
+
+
+
+ FinMME train split (MCQ n = 9,247; Standard n = 1,862).
+ McNemar p ≈ 1.1 × 10⁻¹⁶ (Gemini-3) and p ≈ 3.0 × 10⁻⁶ (Qwen3.6-FP8).
+ †Standard Δ for Qwen is within noise (CI ≈ ±10 pp); treat as exploratory.
+
+
+
+
Key Findings
+
+
+7.7 pp over same-model zero-shot on FinMME (1,250 matched samples, statistically significant at p < 10⁻¹⁵).
+
Multi-select MCQ accuracy improved from 30.2% → 53.5% (+23.3 pp) by giving each agent a dedicated multi-select reasoning path.
+
Verifier adds value: removing the VerifierAgent drops accuracy by ~1.2 pp; revised samples achieve 55.6% exact accuracy vs 68.2% on confirmed samples, confirming the verifier correctly targets harder items.
# Generate MEPs for a dataset split
+uv run --env-file .env -m agentfinvqa.runner.run_generate_meps \
+ --dataset finmme \
+ --split test \
+ --n 200 \
+ --config openai_gemini \
+ --workers 8 \
+ --out meps/
+
+# Evaluate and summarize
+uv run -m agentfinvqa.eval.summarize --mep-dir meps/
+
+# Launch interactive dashboard
+uv run streamlit run src/agentfinvqa/eval/dashboard.py -- --mep-dir meps/
+
+
+
+
+
+
BibTeX
+
If you use AgentFinVQA in your research, please cite:
+
+
+
+
@article{narayanan2026agentfinvqa,
+ title = {AgentFinVQA: A Multi-Agent Framework for Financial Chart Visual Question Answering},
+ author = {Narayanan, Aravind and Raza, Shaina},
+ journal = {arXiv preprint arXiv:XXXX.XXXXX},
+ year = {2026}
+}
+
+
+
+
+
+
Acknowledgements
+
+ Resources used in preparing this research were provided, in part, by the Province of Ontario
+ and the Government of Canada through CIFAR, as well as companies sponsoring the Vector Institute
+ (vectorinstitute.ai/#partners).
+
+
+ This research was funded by the European Union's Horizon Europe research and innovation programme
+ under the AIXPERT project
+ (Grant Agreement No. 101214389), which aims to develop an agentic, multi-layered, GenAI-powered
+ framework for creating explainable, accountable, and transparent AI systems.
+
+
+
+
+
+
+
+
diff --git a/docs/index.md b/docs/index.md
deleted file mode 100644
index 019f42f..0000000
--- a/docs/index.md
+++ /dev/null
@@ -1,4 +0,0 @@
-# AgentFinVQA
-
-Please check out the user guide page for more detailed information on using this project.
-For existing projects, the [user guide](user_guide.md) can be followed to get started.
diff --git a/docs/main.js b/docs/main.js
new file mode 100644
index 0000000..de1df53
--- /dev/null
+++ b/docs/main.js
@@ -0,0 +1,51 @@
+"use strict";
+
+// Mobile nav toggle
+const navToggle = document.getElementById("navToggle");
+const nav = document.getElementById("nav");
+if (navToggle && nav) {
+ navToggle.addEventListener("click", () => {
+ const open = nav.classList.toggle("open");
+ navToggle.setAttribute("aria-expanded", String(open));
+ });
+ // Close on nav link click
+ nav.querySelectorAll("a").forEach((a) => {
+ a.addEventListener("click", () => {
+ nav.classList.remove("open");
+ navToggle.setAttribute("aria-expanded", "false");
+ });
+ });
+}
+
+// Inline code-copy buttons (data-target points to a
id)
+document.querySelectorAll(".code-copy").forEach((btn) => {
+ btn.addEventListener("click", () => {
+ const target = document.getElementById(btn.dataset.target);
+ if (!target) return;
+ navigator.clipboard.writeText(target.textContent.trim()).then(() => {
+ btn.textContent = "Copied!";
+ setTimeout(() => {
+ btn.innerHTML = copyIconSVG();
+ }, 1800);
+ });
+ });
+});
+
+// BibTeX copy button
+const copyBibtex = document.getElementById("copyBibtex");
+const bibtexBlock = document.getElementById("bibtexBlock");
+const copyStatus = document.getElementById("copyStatus");
+if (copyBibtex && bibtexBlock) {
+ copyBibtex.addEventListener("click", () => {
+ navigator.clipboard.writeText(bibtexBlock.textContent.trim()).then(() => {
+ if (copyStatus) copyStatus.textContent = "Copied to clipboard!";
+ setTimeout(() => { if (copyStatus) copyStatus.textContent = ""; }, 2000);
+ });
+ });
+}
+
+function copyIconSVG() {
+ return ``;
+}
diff --git a/docs/style.css b/docs/style.css
new file mode 100644
index 0000000..6477f65
--- /dev/null
+++ b/docs/style.css
@@ -0,0 +1,221 @@
+/* ========== Reset & base ========== */
+*, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
+
+:root {
+ --color-bg: #ffffff;
+ --color-surface: #f8f9fa;
+ --color-border: #e2e8f0;
+ --color-text: #1a202c;
+ --color-muted: #64748b;
+ --color-primary: #0e3c5c; /* Vector dark blue */
+ --color-accent: #1a9fc4; /* Vector teal */
+ --color-accent-light: #e8f6fb;
+ --color-highlight: #fff8e1;
+ --font-sans: system-ui, -apple-system, "Segoe UI", sans-serif;
+ --font-mono: "JetBrains Mono", "Fira Code", "Cascadia Code", monospace;
+ --radius: 8px;
+ --shadow-sm: 0 1px 3px rgba(0,0,0,.08);
+ --shadow-md: 0 4px 12px rgba(0,0,0,.10);
+ --max-w: 1100px;
+ --nav-h: 56px;
+}
+
+html { scroll-behavior: smooth; scroll-padding-top: calc(var(--nav-h) + 16px); }
+body { font-family: var(--font-sans); color: var(--color-text); background: var(--color-bg); line-height: 1.65; font-size: 16px; }
+a { color: var(--color-accent); text-decoration: none; }
+a:hover { text-decoration: underline; }
+img { max-width: 100%; height: auto; display: block; }
+code { font-family: var(--font-mono); font-size: .88em; background: var(--color-surface); padding: .1em .35em; border-radius: 4px; }
+pre code { background: none; padding: 0; font-size: .85rem; }
+
+/* ========== Layout ========== */
+.container { max-width: var(--max-w); margin: 0 auto; padding: 0 1.5rem; }
+.section { padding: 4rem 0; border-bottom: 1px solid var(--color-border); }
+.section:last-child { border-bottom: none; }
+.section-intro { font-size: 1.05rem; color: var(--color-muted); margin-bottom: 1.75rem; max-width: 70ch; }
+
+/* ========== Topbar ========== */
+.topbar {
+ position: sticky; top: 0; z-index: 100;
+ background: rgba(255,255,255,.95); backdrop-filter: blur(8px);
+ border-bottom: 1px solid var(--color-border);
+ height: var(--nav-h);
+}
+.topbar-inner {
+ max-width: var(--max-w); margin: 0 auto; padding: 0 1.5rem;
+ height: 100%; display: flex; align-items: center; gap: 1rem;
+}
+.topbar-logos { display: flex; align-items: center; gap: .75rem; }
+.logo-img { height: 28px; width: auto; }
+.brand { display: flex; align-items: center; gap: .4rem; font-weight: 700; font-size: 1rem; color: var(--color-primary); margin-left: .5rem; }
+.brand:hover { text-decoration: none; }
+.brand-mark { font-size: .9em; opacity: .7; }
+.nav { display: flex; gap: 1.5rem; margin-left: auto; align-items: center; }
+.nav a { font-size: .9rem; color: var(--color-muted); font-weight: 500; transition: color .15s; }
+.nav a:hover { color: var(--color-accent); text-decoration: none; }
+.nav-toggle { display: none; background: none; border: none; cursor: pointer; padding: .4rem; flex-direction: column; gap: 5px; }
+.nav-toggle span { display: block; width: 22px; height: 2px; background: var(--color-text); border-radius: 2px; transition: .2s; }
+
+/* ========== Hero ========== */
+.hero { background: var(--color-bg); color: var(--color-text); padding: 3rem 0 2.5rem; border-bottom: 1px solid var(--color-border); }
+.hero-content { display: flex; flex-direction: column; gap: 2.5rem; align-items: stretch; }
+.hero-header { text-align: center; max-width: 820px; margin: 0 auto; width: 100%; }
+.hero-intro { text-align: center; max-width: 820px; margin: 0 auto; width: 100%; }
+.title { font-size: clamp(1.5rem, 3vw, 2.25rem); font-weight: 800; line-height: 1.25; margin-bottom: 1rem; color: var(--color-primary); }
+.authors { font-size: .95rem; margin-bottom: .4rem; color: var(--color-text); }
+.authors a { color: var(--color-accent); }
+.authors a:hover { color: var(--color-primary); }
+.affiliations { font-size: .85rem; color: var(--color-muted); margin-bottom: 1.25rem; }
+.lead { font-size: 1rem; color: var(--color-muted); margin-bottom: 1.25rem; line-height: 1.7; }
+.hero-attribution { font-size: .78rem; color: var(--color-muted); margin-bottom: 0; }
+.hero-attribution a { color: var(--color-accent); text-decoration: underline; }
+.lead strong { color: var(--color-text); }
+.hero-header .button-row { justify-content: center; margin-bottom: 1rem; }
+.hero-intro .badge-row { justify-content: center; }
+.hero-figure { width: 100%; max-width: var(--max-w); margin: 0 auto; }
+.hero-figure figure { background: var(--color-surface); border: 1px solid var(--color-border); border-radius: var(--radius); overflow: hidden; box-shadow: var(--shadow-sm); }
+.hero-figure .teaser-img { display: block; width: 100%; height: auto; vertical-align: middle; }
+.hero-figure figcaption { padding: .85rem 1.1rem; font-size: .82rem; color: var(--color-muted); line-height: 1.55; text-align: left; background: var(--color-bg); }
+.hero-figure figcaption code { font-size: .78rem; background: var(--color-surface); padding: .1rem .35rem; border-radius: 4px; }
+
+/* ========== Buttons ========== */
+.button-row { display: flex; flex-wrap: wrap; gap: .6rem; margin-bottom: 1.5rem; }
+.btn {
+ display: inline-flex; align-items: center; gap: .4rem;
+ padding: .45rem 1rem; border-radius: 6px; font-size: .88rem; font-weight: 600;
+ border: 1.5px solid var(--color-border); color: var(--color-text); background: var(--color-bg);
+ transition: background .15s, border-color .15s; cursor: pointer; text-decoration: none;
+}
+.btn:hover { background: var(--color-surface); border-color: var(--color-accent); text-decoration: none; color: var(--color-primary); }
+.btn-primary { background: var(--color-accent); border-color: var(--color-accent); color: #fff; }
+.btn-primary:hover { background: #1589ab; border-color: #1589ab; color: #fff; }
+.btn-small { padding: .3rem .7rem; font-size: .8rem; }
+.btn-copy { border: 1.5px solid var(--color-border); color: var(--color-muted); background: var(--color-surface); }
+.btn-copy:hover { background: var(--color-accent-light); color: var(--color-accent); border-color: var(--color-accent); }
+.btn-icon { width: 16px; height: 16px; flex-shrink: 0; }
+.btn-icon-fa { font-size: 1rem; line-height: 1; }
+
+/* ========== Badges / pills ========== */
+.badge-row { display: flex; flex-wrap: wrap; gap: .5rem; }
+.pill { background: var(--color-accent-light); border: 1px solid #b3e0f0; color: var(--color-primary); border-radius: 20px; padding: .3rem .85rem; font-size: .8rem; font-weight: 500; }
+.pill-small { background: var(--color-accent-light); color: var(--color-accent); border-radius: 20px; padding: .2rem .65rem; font-size: .78rem; font-weight: 600; }
+
+/* ========== Headings ========== */
+h2 { font-size: 1.65rem; font-weight: 800; color: var(--color-primary); margin-bottom: 1.25rem; }
+.h3 { font-size: 1.1rem; font-weight: 700; color: var(--color-text); margin-bottom: .75rem; }
+
+/* ========== Abstract box ========== */
+.abstract-box { background: var(--color-surface); border-left: 4px solid var(--color-accent); border-radius: 0 var(--radius) var(--radius) 0; padding: 1.5rem 1.75rem; }
+.abstract-box p + p { margin-top: .9rem; }
+
+/* ========== Pipeline ========== */
+.pipeline { display: flex; flex-direction: column; align-items: center; gap: 0; max-width: 640px; margin: 0 auto; }
+.pipeline-step {
+ display: flex; align-items: flex-start; gap: 1rem;
+ background: var(--color-surface); border: 1px solid var(--color-border);
+ border-radius: var(--radius); padding: 1rem 1.25rem; width: 100%;
+}
+.pipeline-step--output { background: var(--color-accent-light); border-color: var(--color-accent); }
+.step-number {
+ flex-shrink: 0; width: 32px; height: 32px; border-radius: 50%;
+ background: var(--color-accent); color: #fff; font-weight: 800; font-size: .9rem;
+ display: flex; align-items: center; justify-content: center;
+}
+.pipeline-step--output .step-number { background: var(--color-primary); }
+.step-body h3 { font-size: 1rem; font-weight: 700; margin-bottom: .3rem; }
+.step-body p { font-size: .9rem; color: var(--color-muted); }
+.pipeline-arrow { color: var(--color-accent); font-size: 1.4rem; padding: .2rem 0 .2rem 15px; }
+
+/* ========== Two-col grid ========== */
+.two-col { display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; align-items: start; }
+.grid-2 { display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; }
+.grid-4 { display: grid; grid-template-columns: repeat(4, 1fr); gap: 1rem; margin-bottom: 2rem; }
+
+/* ========== Stat ========== */
+.stat { text-align: center; padding: 1rem; background: var(--color-surface); border-radius: var(--radius); border: 1px solid var(--color-border); }
+.stat-value { font-size: 2rem; font-weight: 800; color: var(--color-primary); line-height: 1; }
+.stat-label { font-size: .8rem; color: var(--color-muted); margin-top: .3rem; }
+
+/* ========== Card ========== */
+.card { background: var(--color-surface); border: 1px solid var(--color-border); border-radius: var(--radius); padding: 1.5rem; }
+.info-box { background: var(--color-accent-light); border: 1px solid #b3e0f0; border-radius: var(--radius); padding: 1.25rem 1.5rem; }
+.info-box .h3 { color: var(--color-primary); }
+
+/* ========== Lists ========== */
+.bullets { list-style: none; padding: 0; }
+.bullets li { padding: .35rem 0 .35rem 1.4rem; position: relative; font-size: .95rem; }
+.bullets li::before { content: "→"; position: absolute; left: 0; color: var(--color-accent); font-weight: 700; }
+.compact { list-style: disc; padding-left: 1.2rem; font-size: .9rem; }
+.compact li { padding: .15rem 0; }
+ol.compact { list-style: decimal; padding-left: 1.4rem; }
+ol.compact li { padding: .2rem 0; font-size: .9rem; }
+
+/* ========== Table ========== */
+.table-wrap { overflow-x: auto; border-radius: var(--radius); border: 1px solid var(--color-border); }
+.table { width: 100%; border-collapse: collapse; font-size: .9rem; }
+.table th { background: var(--color-primary); color: #fff; padding: .65rem 1rem; text-align: left; font-weight: 600; white-space: nowrap; }
+.table td { padding: .6rem 1rem; border-bottom: 1px solid var(--color-border); }
+.table tbody tr:last-child td { border-bottom: none; }
+.table tbody tr:hover td { background: var(--color-surface); }
+.table .highlight td { background: #f0f9ff; font-weight: 600; }
+.table .delta-row td { background: #f0fdf4; font-size: .85rem; color: #166534; border-bottom: 2px solid var(--color-border); }
+.table .center { text-align: center; }
+.metric-desc { font-weight: 400; font-size: .78rem; color: rgba(255,255,255,.75); }
+.table-note { font-size: .82rem; color: var(--color-muted); margin-top: .75rem; }
+
+/* ========== Figure placeholder ========== */
+.figure-placeholder {
+ background: rgba(255,255,255,.08); border: 2px dashed rgba(255,255,255,.2);
+ border-radius: var(--radius); padding: 3rem 1rem;
+ display: flex; flex-direction: column; align-items: center; justify-content: center;
+ gap: .5rem; color: rgba(255,255,255,.5); text-align: center; min-height: 220px;
+}
+.figure-placeholder span { font-size: 1rem; font-weight: 600; }
+.figure-placeholder p { font-size: .85rem; }
+
+/* ========== Code cards ========== */
+.code-section { margin-bottom: 2rem; }
+.code-card { position: relative; background: #1e293b; border-radius: var(--radius); overflow: hidden; }
+.code-card pre { padding: 1.25rem 1.5rem; overflow-x: auto; }
+.code-card code { color: #e2e8f0; font-size: .85rem; }
+.code-copy {
+ position: absolute; top: .6rem; right: .6rem;
+ background: rgba(255,255,255,.1); border: 1px solid rgba(255,255,255,.15);
+ color: rgba(255,255,255,.7); border-radius: 5px; padding: .3rem .5rem;
+ cursor: pointer; display: flex; align-items: center; transition: background .15s;
+}
+.code-copy:hover { background: rgba(255,255,255,.2); color: #fff; }
+
+/* ========== BibTeX ========== */
+.bibtex-section { position: relative; margin-top: 1rem; }
+.bibtex { background: #1e293b; color: #e2e8f0; border-radius: var(--radius); padding: 1.25rem 1.5rem; overflow-x: auto; font-size: .85rem; margin-top: .75rem; }
+.copy-status { font-size: .82rem; color: var(--color-accent); height: 1.2em; margin-top: .4rem; }
+
+/* ========== Footer ========== */
+.footer { background: var(--color-primary); color: rgba(255,255,255,.7); padding: 2rem 0; margin-top: 2rem; }
+.footer-inner { display: flex; align-items: center; gap: 1.5rem; }
+.footer-logo { height: 28px; filter: brightness(0) invert(1); opacity: .85; }
+.footer-logo--color { filter: none; opacity: .9; }
+.footer-text { font-size: .88rem; }
+.footer-text a { color: rgba(255,255,255,.7); }
+.footer-text a:hover { color: #fff; }
+
+/* ========== Misc ========== */
+.muted { color: var(--color-muted); }
+.small { font-size: .85rem; }
+sup { font-size: .7em; vertical-align: super; }
+
+/* ========== Responsive ========== */
+@media (max-width: 768px) {
+ .two-col { grid-template-columns: 1fr; }
+ .grid-4 { grid-template-columns: 1fr 1fr; }
+ .nav { display: none; flex-direction: column; position: absolute; top: var(--nav-h); left: 0; right: 0; background: #fff; border-bottom: 1px solid var(--color-border); padding: 1rem 1.5rem; gap: .75rem; z-index: 99; }
+ .nav.open { display: flex; }
+ .nav-toggle { display: flex; margin-left: auto; }
+ .footer-inner { flex-direction: column; text-align: center; gap: .75rem; }
+}
+
+@media (max-width: 480px) {
+ .grid-4 { grid-template-columns: 1fr; }
+ .grid-2 { grid-template-columns: 1fr; }
+}
diff --git a/docs/user_guide.md b/docs/user_guide.md
deleted file mode 100644
index fb42741..0000000
--- a/docs/user_guide.md
+++ /dev/null
@@ -1,158 +0,0 @@
-# User Guide
-
-## pyproject.toml file and dependency management
-
-If your project doesn't have a pyproject.toml file, simply copy the one from the
-template and update file according to your project.
-
-For managing dependencies, this template makes use of [uv](https://docs.astral.sh/uv/),
-which according to some [benchmarks](https://github.com/astral-sh/uv/blob/main/BENCHMARKS.md)
-is faster than alternative like Poetry (which our original AI Engineering Template
-makes use of).
-
-Hence, be sure to install uv in order to to setup the development virtual environment.
-Instructions for installing uv can be found [here](https://docs.astral.sh/uv/getting-started/installation/).
-Note that uv supports [optional dependency groups](https://docs.astral.sh/uv/concepts/projects/dependencies/#dependency-groups)
-which helps to manage dependencies for different parts of development such as `documentation`,
-`testing`, etc. The core dependencies are installed using the command:
-
-```bash
-uv sync
-```
-
-Additional dependency groups can be installed using the `--group` flag followed
-by the group name. For example:
-
-```bash
-uv sync --all-extras --group docs --group test
-```
-
-!!! important "mypy configuration options"
- By default, the `mypy` configuration in the `pyproject.toml` disallows subclassing
- the `Any` type - `allow_subclassing_any = false`. In cases where the type checker
- is not able to determine the types of objects in some external library (e.g. `PyTorch`),
- it will treat them as `Any` and raise errors. If your codebase has many of such
- cases, you can set `allow_subclassing_any = true` in the `mypy` configuration or
- remove it entirely to use the default value (which is `true`). For example, in
- a `PyTorch` project, subclassing `nn.Module` will raise errors if `allow_subclassing_any`
- is set to `false`.
-
-
-## pre-commit
-
-You can use [pre-commit](https://pre-commit.com/) to run pre-commit hooks (code checks,
-liniting, etc.) when you run `git commit` and commit your code. Simply copy the
-`.pre-commit-config.yaml` file to the root of the repository and install the test
-dependencies which installs pre-commit. Then run:
-
-```bash
-pre-commit install
-```
-
-If you prefer to not enforce using pre-commit every time you run `git commit`,
-you will have to run `pre-commit run --all-files` from the command line before you
-commit your code.
-
-!!! important "hook configuration"
- Some of the pre-commit hooks use [supported hooks](https://pre-commit.com/hooks.html)
- from the web.
-
- For some others, they are locally installed and hence use the python virtual environment
- locally. If `language` is set to `python`, each time the hook is installed, a separate
- python virtual environment is created and you can specify dependencies needed using
- `additional_dependencies`.
-
- If `language` is set to `system`, the activated python virtual environment is used and
- and hence you have to ensure that the required dependencies and their versions are
- correctly installed.
-
- ```yaml
- - repo: local
- hooks:
- - id: pytest
- name: pytest
- entry: python3 -m pytest -m "not integration_test"
- language: python/system # set according to your project needs
- ```
-
-!!! warning "typos"
- The [typos](https://github.com/crate-ci/typos) pre-commit hook is used to check for
- common spelling mistakes in the codebase. While useful, it may require some
- configuration to ignore certain words or phrases that are not typos. You can
- [configure the typos hook](https://github.com/crate-ci/typos/blob/master/docs/reference.md)
- in the `pyproject.toml` file. In a large codebase, it may be useful to disable
- the typos hook and only run it occasionally on the entire codebase.
-
-
-### pre-commit ci
-
-Instead of fixing pre-commit errors manually, a CI to fix them as well as update
-pre-commit hooks periodically can be enabled for your repository. Please check
-[pre-commit.ci](https://pre-commit.ci/) and add your repository. The configuration for
-``pre-commit.ci`` can be added to the ``.pre-commit-config.yaml`` file.
-
-
-## documentation
-
-If your project doesn't have documentation, copy the directory named `docs` to the root
-directory of your repository. This template uses [MkDocs](https://www.mkdocs.org/) with
-the [Material for MkDocs](https://squidfunk.github.io/mkdocs-material/) theme.
-
-In order to build the documentation, install the documentation dependencies as mentioned
-in the previous section, then run the command:
-
-```bash
-mkdocs build
-```
-
-If you're making changes to the docs, and want to serve them locally on your machine,
-then you can use this command instead:
-
-```bash
-mkdocs serve
-```
-
-The above will launch the docs locally on `http://127.0.0.1:8000`, which you can
-enter into your browser of choice. Conveniently, this process also watches for any
-changes you make to the docs and will update them as they occur.
-
-You can configure the documentation by updating the `mkdocs.yml` file at the root of
-your repository. The markdown files in the `docs` directory can be updated to reflect
-the project's documentation.
-
-### GitHub Pages Setup
-
-To serve your documentation on GitHub Pages:
-
-1. Go to your repository's **Settings** tab
-2. Navigate to **Pages** in the left sidebar
-3. Under **Source**, select **Deploy from a branch**
-4. Choose the **gh-pages** branch and **/ (root)** folder
-5. Click **Save**
-
-The documentation will be automatically built and deployed to GitHub Pages whenever you push changes to the main branch (thanks to the `.github/workflows/docs.yml` workflow). Your site will be available at `https://[username].github.io/[repository-name]/`.
-
-
-## github actions
-
-The template consists of some github action continuous integration workflows that you
-can add to your repository.
-
-The available workflows are:
-
-- [code checks](https://github.com/VectorInstitute/aieng-template/blob/main/.github/workflows/code_checks.yml): Static code analysis, code formatting and unit tests
-- [documentation](https://github.com/VectorInstitute/aieng-template/blob/main/.github/workflows/docs.yml): Project documentation including example API reference
-- [integration tests](https://github.com/VectorInstitute/aieng-template/blob/main/.github/workflows/integration_tests.yml): Integration tests
-- [publish](https://github.com/VectorInstitute/aieng-template/blob/main/.github/workflows/publish.yml):
-Publishing python package to PyPI. Create a `PYPI_API_TOKEN` and add it to the
-repository's actions [secret variables](https://docs.github.com/en/actions/security-guides/using-secrets-in-github-actions)
-in order to publish PyPI packages when new software releases are created on Github.
-
-The test workflows also compute coverage and upload code coverage metrics to
-[codecov.io](https://app.codecov.io/gh/VectorInstitute/aieng-template). Create a
-`CODECOV_TOKEN` and add it to the repository's actions [secret variables](https://docs.github.com/en/actions/security-guides/using-secrets-in-github-actions).
-
-!!! warning "codecov"
- The [codecov](https://app.codecov.io/github/VectorInstitute) tool is subscribed under the free tier
- which makes it usable only for public open-source repos. Hence, if you would like to develop in a
- private repo, it is recommended to remove the codecov actions from the github workflow files.
diff --git a/mkdocs.yml b/mkdocs.yml
index f9ba705..94c0c35 100644
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -7,60 +7,19 @@ extra:
link: 404.html
- icon: fontawesome/brands/github
link: https://github.com/VectorInstitute/AgentFinVQA
-markdown_extensions:
- - attr_list
- - admonition
- - md_in_html
- - pymdownx.highlight:
- anchor_linenums: true
- line_spans: __span
- pygments_lang_class: true
- - pymdownx.inlinehilite
- - pymdownx.details
- - pymdownx.snippets
- - pymdownx.superfences
- - pymdownx.emoji:
- emoji_index: !!python/name:material.extensions.emoji.twemoji
- emoji_generator: !!python/name:material.extensions.emoji.to_svg
- - toc:
- permalink: true
-nav:
- - Home: index.md
- - User Guide: user_guide.md
- - API Reference: api.md
-plugins:
- - search
- - mkdocstrings:
- handlers:
- python:
- options:
- docstring_style: numpy
- members_order: source
- separate_signature: true
- show_overloads: true
- show_submodules: true
- show_root_heading: false
- show_root_full_path: true
- show_root_toc_entry: false
- show_symbol_type_heading: true
- show_symbol_type_toc: true
repo_url: https://github.com/VectorInstitute/AgentFinVQA
repo_name: VectorInstitute/AgentFinVQA
site_name: AgentFinVQA
+plugins:
+ - search
theme:
custom_dir: docs/overrides
favicon: assets/favicon-48x48.svg
features:
- - content.code.annotate
- - content.code.copy
- navigation.footer
- - navigation.indexes
- - navigation.instant
- - navigation.tabs
- navigation.top
- search.suggest
- search.highlight
- - toc.follow
icon:
repo: fontawesome/brands/github
logo: assets/vector-logo.svg
+ 
+ 
+