ticket #104 attempt 1

devplacepy/main.py

@@ -12,6 +12,7 @@ from fastapi import FastAPI, Request
 from fastapi.responses import HTMLResponse, RedirectResponse
 from fastapi.staticfiles import StaticFiles
 from fastapi.exceptions import RequestValidationError
+from starlette.middleware.gzip import GZipMiddleware
 from devplacepy.config import (
     STATIC_DIR,
     STATIC_VERSION,
@@ -609,6 +610,10 @@ async def response_timing(request: Request, call_next):
     response.headers["X-Response-Time"] = f"{(time.perf_counter() - start) * 1000:.1f}ms"
     return response
 
+
+app.add_middleware(GZipMiddleware, minimum_size=512, compresslevel=5)
+
+
 _home_cache = TTLCache(ttl=int(os.environ.get("DEVPLACE_HOME_CACHE_TTL", "60")), max_size=4)
 
 

devplacepy/services/jobs/deepsearch/crawl.py

@@ -22,6 +22,8 @@ from .pdf import MAX_PDF_BYTES, extract_pdf_text, is_pdf
 
 logger = logging.getLogger(__name__)
 
+_pw_lock = asyncio.Lock()
+
 RSEARCH_URL = "https://rsearch.app.molodetz.nl"
 RSEARCH_TIMEOUT_SECONDS = 45.0
 FETCH_TIMEOUT_SECONDS = 20.0
@@ -172,13 +174,7 @@ async def _render_with_playwright(
 ) -> tuple[str, str, int, list[tuple[str, str]]]:
     from playwright.async_api import async_playwright
 
-    own_browser = browser is None
-    if own_browser:
-        pw = await async_playwright().__aenter__()
-        browser = await pw.chromium.launch(
-            headless=True, args=["--no-sandbox", "--disable-dev-shm-usage"]
-        )
-    try:
+    async def _render(browser) -> tuple[str, str, int, list[tuple[str, str]]]:
         context = await browser.new_context(user_agent=USER_AGENT)
         page = await context.new_page()
         response = await page.goto(url, wait_until="load", timeout=30000)
@@ -189,10 +185,18 @@ async def _render_with_playwright(
         await context.close()
         extracted = extract_html(content, base_url=url)
         return extracted.title, extracted.text, status, extracted.links
-    finally:
-        if own_browser:
-            await browser.close()
-            await pw.__aexit__(None, None, None)
+
+    if browser is None:
+        async with async_playwright() as pw:
+            browser = await pw.chromium.launch(
+                headless=True, args=["--no-sandbox", "--disable-dev-shm-usage"]
+            )
+            try:
+                return await _render(browser)
+            finally:
+                await browser.close()
+    else:
+        return await _render(browser)
 
 
 async def fetch_page(url: str, depth: int, browser=None) -> CrawledPage | None:
@@ -242,18 +246,20 @@ async def fetch_page(url: str, depth: int, browser=None) -> CrawledPage | None:
         except (LookupError, ValueError) as exc:
             logger.info("deepsearch decode failed for %s: %s", url, exc)
     if len(text) < MIN_PAGE_CHARS:
-        try:
-            r_title, r_text, r_status, r_links = await _render_with_playwright(url, browser)
-            if len(r_text) > len(text):
-                title, text, status, source, links = (
-                    r_title or title,
-                    r_text,
-                    r_status or status,
-                    "playwright",
-                    r_links,
-                )
-        except Exception as exc:
-            logger.info("deepsearch render failed for %s: %s", url, exc)
+        async with _pw_lock:
+            try:
+                r_title, r_text, r_status, r_links = await _render_with_playwright(url, browser)
+            except Exception as exc:
+                logger.info("deepsearch render failed for %s: %s", url, exc)
+                r_title, r_text, r_status, r_links = "", "", 0, []
+        if len(r_text) > len(text):
+            title, text, status, source, links = (
+                r_title or title,
+                r_text,
+                r_status or status,
+                "playwright",
+                r_links,
+            )
     if len(text) < MIN_PAGE_CHARS:
         return None
     return CrawledPage(
@@ -296,102 +302,99 @@ async def crawl(
     total = min(len(level_candidates), max_pages)
     cancelled = False
 
-    pw = None
     browser = None
-    try:
-        pw = await async_playwright().__aenter__()
-        browser = await pw.chromium.launch(
-            headless=True, args=["--no-sandbox", "--disable-dev-shm-usage"]
-        )
-    except Exception as exc:
-        logger.warning("deepsearch playwright launch failed, pages will use httpx only: %s", exc)
+    async with async_playwright() as pw:
+        try:
+            browser = await pw.chromium.launch(
+                headless=True, args=["--no-sandbox", "--disable-dev-shm-usage"]
+            )
+        except Exception as exc:
+            logger.warning("deepsearch playwright launch failed, pages will use httpx only: %s", exc)
 
-    try:
-        for level in range(max(1, depth)):
-            if cancelled or fetched >= max_pages or not level_candidates:
-                break
-            next_candidates: list[dict] = []
-            for start in range(0, len(level_candidates), CRAWL_CONCURRENCY):
-                if fetched >= max_pages:
+        try:
+            for level in range(max(1, depth)):
+                if cancelled or fetched >= max_pages or not level_candidates:
                     break
-                if await should_stop():
-                    emit({"type": "stage", "stage": "cancelled", "message": "Crawl cancelled"})
-                    cancelled = True
-                    break
-                batch = level_candidates[start : start + CRAWL_CONCURRENCY][: max_pages - fetched]
-                for candidate in batch:
-                    emit(
-                        {
-                            "type": "progress",
-                            "done": fetched,
-                            "total": total,
-                            "url": candidate["url"],
-                            "depth": level,
-                            "message": f"Reading {candidate['url']}",
-                        }
-                    )
-                    if is_cached(candidate["url"]):
-                        emit({"type": "page_cached", "url": candidate["url"], "reason": "seen in a prior run"})
-                fetch_start = time.perf_counter()
-                results = await asyncio.gather(
-                    *(_resolve_candidate(candidate, level, browser) for candidate in batch),
-                    return_exceptions=True,
-                )
-                elapsed_ms = int((time.perf_counter() - fetch_start) * 1000)
-                for candidate, page in zip(batch, results):
-                    url = candidate["url"]
-                    if isinstance(page, BaseException):
-                        logger.info("deepsearch fetch crashed for %s: %s", url, page)
-                        page = None
-                    if page is None:
-                        emit(
-                            {
-                                "type": "page_skipped",
-                                "url": url,
-                                "reason": "no readable content",
-                                "elapsed_ms": elapsed_ms,
-                            }
-                        )
-                        continue
+                next_candidates: list[dict] = []
+                for start in range(0, len(level_candidates), CRAWL_CONCURRENCY):
                     if fetched >= max_pages:
                         break
-                    digest = content_hash(page.text)
-                    if digest in outcome.seen_hashes:
+                    if await should_stop():
+                        emit({"type": "stage", "stage": "cancelled", "message": "Crawl cancelled"})
+                        cancelled = True
+                        break
+                    batch = level_candidates[start : start + CRAWL_CONCURRENCY][: max_pages - fetched]
+                    for candidate in batch:
                         emit(
                             {
-                                "type": "page_duplicate",
-                                "url": url,
-                                "reason": "duplicate content",
-                                "elapsed_ms": elapsed_ms,
+                                "type": "progress",
+                                "done": fetched,
+                                "total": total,
+                                "url": candidate["url"],
+                                "depth": level,
+                                "message": f"Reading {candidate['url']}",
                             }
                         )
-                        continue
-                    outcome.seen_hashes.add(digest)
-                    outcome.pages.append(page)
-                    fetched += 1
-                    emit(
-                        {
-                            "type": "page_loaded",
-                            "url": page.url,
-                            "title": page.title,
-                            "source": page.source,
-                            "depth": level,
-                            "render": page.source == "playwright",
-                            "elapsed_ms": elapsed_ms,
-                            "done": fetched,
-                            "total": total,
-                        }
+                        if is_cached(candidate["url"]):
+                            emit({"type": "page_cached", "url": candidate["url"], "reason": "seen in a prior run"})
+                    fetch_start = time.perf_counter()
+                    results = await asyncio.gather(
+                        *(_resolve_candidate(candidate, level, browser) for candidate in batch),
+                        return_exceptions=True,
                     )
-                    if level + 1 < depth:
-                        for link in relevant_links(page.links, query, LINKS_PER_PAGE):
-                            if link not in seen_urls:
-                                seen_urls.add(link)
-                                next_candidates.append({"url": link})
-            level_candidates = next_candidates
-            total = min(total + len(next_candidates), max_pages)
-    finally:
-        if browser is not None:
-            await browser.close()
-        if pw is not None:
-            await pw.__aexit__(None, None, None)
+                    elapsed_ms = int((time.perf_counter() - fetch_start) * 1000)
+                    for candidate, page in zip(batch, results):
+                        url = candidate["url"]
+                        if isinstance(page, BaseException):
+                            logger.info("deepsearch fetch crashed for %s: %s", url, page)
+                            page = None
+                        if page is None:
+                            emit(
+                                {
+                                    "type": "page_skipped",
+                                    "url": url,
+                                    "reason": "no readable content",
+                                    "elapsed_ms": elapsed_ms,
+                                }
+                            )
+                            continue
+                        if fetched >= max_pages:
+                            break
+                        digest = content_hash(page.text)
+                        if digest in outcome.seen_hashes:
+                            emit(
+                                {
+                                    "type": "page_duplicate",
+                                    "url": url,
+                                    "reason": "duplicate content",
+                                    "elapsed_ms": elapsed_ms,
+                                }
+                            )
+                            continue
+                        outcome.seen_hashes.add(digest)
+                        outcome.pages.append(page)
+                        fetched += 1
+                        emit(
+                            {
+                                "type": "page_loaded",
+                                "url": page.url,
+                                "title": page.title,
+                                "source": page.source,
+                                "depth": level,
+                                "render": page.source == "playwright",
+                                "elapsed_ms": elapsed_ms,
+                                "done": fetched,
+                                "total": total,
+                            }
+                        )
+                        if level + 1 < depth:
+                            for link in relevant_links(page.links, query, LINKS_PER_PAGE):
+                                if link not in seen_urls:
+                                    seen_urls.add(link)
+                                    next_candidates.append({"url": link})
+                level_candidates = next_candidates
+                total = min(total + len(next_candidates), max_pages)
+        finally:
+            if browser is not None:
+                await browser.close()
     return outcome

devplacepy/templates/docs/architecture-backend.html

@@ -11,7 +11,7 @@ How a request flows through middleware to a router, how the database layer is bu
 
 ## Middleware
 
-Six HTTP middlewares run as a stack around every request, listed outermost first. `response_timing` is the outermost and `refresh_db_snapshot` the innermost. Compression is handled by nginx in production; the Python application does not compress responses itself.
+Seven HTTP middlewares run as a stack around every request, listed outermost first. `response_timing` is the outermost, `refresh_db_snapshot` the innermost, and a `GZipMiddleware` (responses over 512 bytes) wraps the whole stack on top:
 
 | Middleware (outermost first) | Responsibility |
 |------------|----------------|

pyproject.toml

@@ -4,8 +4,7 @@ version = "1.0.0"
 description = "DevPlace - The Developer Social Network"
 requires-python = ">=3.12"
 dependencies = [
-    "fastapi>=0.110.0",
-    "starlette>=0.37.0",
+    "fastapi",
     "uvicorn[standard]",
     "jinja2",
     "python-multipart",