chore: add brand-styled logo + screenshot assets for adk-docs PR

.gitignore

@@ -19,3 +19,8 @@ venv/
 .idea/
 .vscode/
 .DS_Store
+
+# Brand asset working files (downloaded by scripts/render_screenshot.py at runtime)
+docs/assets/FTSystemMono-*.ttf
+docs/assets/parallel-symbol-*.png
+docs/assets/parallel-logo-*.png

README.md

@@ -1,12 +1,20 @@
+<p align="center">
+  <img src="docs/assets/parallel.png" alt="Parallel" width="120" height="120" />
+</p>
+
 # parallel-google-adk
 
+[![PyPI](https://img.shields.io/pypi/v/parallel-google-adk.svg)](https://pypi.org/project/parallel-google-adk/)
+[![Python](https://img.shields.io/pypi/pyversions/parallel-google-adk.svg)](https://pypi.org/project/parallel-google-adk/)
+[![License](https://img.shields.io/pypi/l/parallel-google-adk.svg)](LICENSE)
+
 Google [Agent Development Kit](https://adk.dev) (ADK) tools and plugin for [Parallel](https://parallel.ai) — grounded web search, clean extraction, and cited deep research with structured output.
 
 ## Install
 
 ```bash
-pip install parallel-google-adk
-export PARALLEL_API_KEY=your-key-here  # get one at https://platform.parallel.ai
+uv add parallel-google-adk           # or: pip install parallel-google-adk
+export PARALLEL_API_KEY=your-key-here # get one at https://platform.parallel.ai
 ```
 
 ## Quickstart
@@ -72,7 +80,9 @@ This package is the typed-FunctionTool path for users who want fine-grained sche
 
 ## Examples
 
-See [`examples/research_agent.py`](examples/research_agent.py) for a runnable demo.
+See [`examples/research_agent.py`](examples/research_agent.py) for a runnable demo. A run looks like this:
+
+![Research agent run](docs/assets/research_agent.png)
 
 ## Development
 

docs/screenshot.txt

@@ -2,15 +2,7 @@ $ python research_agent.py
 
 >>> What does Parallel's Search API do? Cite parallel.ai in one sentence.
 
-Parallel’s Search API is a tool designed specifically for AI agents and LLMs to perform natural language web searches and retrieve structured, token-efficient data. Unlike traditional search engines that rely on keywords, it processes a "natural language objective" to return ranked URLs with extended excerpts optimized for model consumption, reducing hallucinations and improving accuracy in complex research tasks ([parallel.ai](https://docs.parallel.ai/search/search-quickstart)).
-
-According to the developer documentation: "The Search API takes a natural language objective and returns relevant excerpts optimized for LLMs, replacing multiple keyword searches with a single call for broad or complex queries" ([parallel.ai](https://docs.parallel.ai/search/search-quickstart)).
-
-### Key Features
-*   **Search Modes:** Offers specialized modes including **one-shot** (comprehensive), **agentic** (concise for multi-step reasoning), and **fast** (latency-sensitive) ([parallel.ai](https://docs.parallel.ai/search/modes)).
-*   **LLM Optimization:** Excerpts are curated to fit efficiently into a model's context window, lowering token costs while maintaining high relevance ([parallel.ai](https://parallel.ai/)).
-*   **Accuracy Benchmarks:** Parallel claims its enterprise deep research API achieves up to 48% accuracy on complex benchmarks, significantly outperforming the native browsing capabilities of standard LLMs ([parallel.ai](https://parallel.ai/)).
-*   **Evidence-Based:** Every output provides verifiability and provenance, ensuring AI agents can cite their sources accurately ([parallel.ai](https://parallel.ai/)).
+Parallel's Search API is a web search engine designed specifically for AI agents that streamlines the search, scrape, and extraction pipeline into a single API call using declarative semantic search to provide token-efficient results ([parallel.ai](https://parallel.ai/products/search)).
 
 --- Parallel call trace (1 calls) ---
-  {"tool": "_web_search", "latency_s": 1.4036233751103282, "citation_count": 5}
+  {"tool": "web_search", "latency_s": 2.6487630419433117, "citation_count": 5}

scripts/render_screenshot.py

@@ -0,0 +1,173 @@
+"""Render docs/screenshot.txt to a brand-styled PNG for the adk-docs PR.
+
+Reads docs/screenshot.txt, lays it out in FT System Mono on Off-white using
+Parallel's brand palette (per assets.parallel.ai/llms.txt style guide), and
+writes docs/assets/research_agent.png.
+
+Fonts and intermediate brand assets are downloaded on demand from
+assets.parallel.ai (gitignored). The committed outputs are docs/assets/parallel.png
+(the symbol logo) and docs/assets/research_agent.png (the rendered screenshot).
+
+Run:
+    .venv/bin/python scripts/render_screenshot.py
+"""
+
+from __future__ import annotations
+
+import sys
+import urllib.request
+from pathlib import Path
+
+from PIL import Image, ImageDraw, ImageFont
+
+ROOT = Path(__file__).resolve().parent.parent
+SRC = ROOT / "docs" / "screenshot.txt"
+DST = ROOT / "docs" / "assets" / "research_agent.png"
+
+# Parallel brand palette (assets.parallel.ai/llms.txt).
+OFF_WHITE = "#fcfcfa"
+INDEX_BLACK = "#1d1b16"
+NEURAL = "#d8d0bf"
+SIGNAL = "#fb631b"
+
+FONT_REGULAR = ROOT / "docs" / "assets" / "FTSystemMono-Regular.ttf"
+FONT_BOLD = ROOT / "docs" / "assets" / "FTSystemMono-Bold.ttf"
+
+FONT_URLS = {
+    FONT_REGULAR: "https://assets.parallel.ai/FTSystemMono-Regular.ttf",
+    FONT_BOLD: "https://assets.parallel.ai/FTSystemMono-Bold.ttf",
+}
+
+
+def ensure_fonts() -> None:
+    """Download the brand fonts if they aren't already cached locally.
+
+    assets.parallel.ai's CDN rejects the default urllib User-Agent (403);
+    pass a browser-style UA so the request goes through.
+    """
+    for path, url in FONT_URLS.items():
+        if path.exists():
+            continue
+        path.parent.mkdir(parents=True, exist_ok=True)
+        print(f"fetching {url}", file=sys.stderr)
+        req = urllib.request.Request(url, headers={"User-Agent": "Mozilla/5.0"})
+        with urllib.request.urlopen(req) as resp, path.open("wb") as fp:
+            fp.write(resp.read())
+
+# Layout.
+WIDTH = 1400
+PADDING_X = 70
+PADDING_Y = 60
+FONT_SIZE = 22
+LINE_HEIGHT = int(FONT_SIZE * 1.55)
+WRAP_WIDTH = 78  # chars
+
+# Window-chrome dots (subtle, no labels).
+DOT_RADIUS = 7
+DOT_GAP = 22
+
+
+def wrap(text: str, width: int) -> list[str]:
+    """Word-wrap a single line of text to a max char width, preserving runs."""
+    if len(text) <= width:
+        return [text]
+    out: list[str] = []
+    line = ""
+    for word in text.split(" "):
+        if not line:
+            line = word
+        elif len(line) + 1 + len(word) <= width:
+            line = f"{line} {word}"
+        else:
+            out.append(line)
+            line = word
+    if line:
+        out.append(line)
+    return out
+
+
+def color_for_line(line: str) -> str:
+    stripped = line.strip()
+    if stripped.startswith("$ "):
+        return SIGNAL
+    if stripped.startswith(">>>"):
+        return SIGNAL
+    if stripped.startswith("---"):
+        return NEURAL
+    if stripped.startswith("{"):
+        return NEURAL
+    return INDEX_BLACK
+
+
+def font_for_line(line: str, regular: ImageFont.FreeTypeFont, bold: ImageFont.FreeTypeFont) -> ImageFont.FreeTypeFont:
+    s = line.strip()
+    if s.startswith("$ ") or s.startswith(">>>"):
+        return bold
+    return regular
+
+
+def main() -> int:
+    if not SRC.exists():
+        print(f"missing: {SRC}", file=sys.stderr)
+        return 1
+    ensure_fonts()
+
+    raw = SRC.read_text().splitlines()
+
+    # Wrap long lines once for layout.
+    lines: list[str] = []
+    for line in raw:
+        if not line:
+            lines.append("")
+        else:
+            lines.extend(wrap(line, WRAP_WIDTH))
+
+    regular = ImageFont.truetype(str(FONT_REGULAR), FONT_SIZE)
+    bold = ImageFont.truetype(str(FONT_BOLD), FONT_SIZE)
+
+    # Compute height: window-chrome row + padding + text + padding.
+    chrome_height = PADDING_Y - 10
+    text_height = max(LINE_HEIGHT * len(lines), 100)
+    height = chrome_height + PADDING_Y + text_height + PADDING_Y
+
+    img = Image.new("RGB", (WIDTH, height), OFF_WHITE)
+    draw = ImageDraw.Draw(img)
+
+    # Window chrome dots.
+    cy = chrome_height // 2 + 18
+    cx = PADDING_X + DOT_RADIUS
+    for i in range(3):
+        x = cx + i * DOT_GAP
+        draw.ellipse(
+            [(x - DOT_RADIUS, cy - DOT_RADIUS), (x + DOT_RADIUS, cy + DOT_RADIUS)],
+            outline=NEURAL,
+            width=2,
+        )
+
+    # Hairline separator between chrome and text.
+    sep_y = chrome_height + PADDING_Y // 2
+    draw.line(
+        [(PADDING_X, sep_y), (WIDTH - PADDING_X, sep_y)],
+        fill=NEURAL,
+        width=1,
+    )
+
+    # Text.
+    y = sep_y + PADDING_Y // 2
+    for line in lines:
+        if line == "":
+            y += LINE_HEIGHT
+            continue
+        font = font_for_line(line, regular, bold)
+        color = color_for_line(line)
+        draw.text((PADDING_X, y), line, font=font, fill=color)
+        y += LINE_HEIGHT
+
+    DST.parent.mkdir(parents=True, exist_ok=True)
+    img.save(DST, optimize=True)
+    print(f"wrote {DST} ({img.size[0]}x{img.size[1]})")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())