Release v0.2.45

* feat: add Keenable keyless web search tool

Keenable is a web-search service (api.keenable.ai). Per AGENTS.md routing,
vendor search services belong in PraisonAI-Tools as standalone tool modules
(mirroring tavily_tool.py / brave_search_tool.py), not inline in the core
praisonaiagents web_search.py aggregator.

Keyless by default via the public endpoint (stdlib urllib, no install);
KEENABLE_API_KEY switches to the authenticated endpoint for higher limits.
Redirects are refused so X-API-Key is never resent to another host; results
are bounded (50 max, 1 MB body cap, title/URL/snippet length caps).

Auto-discovered by praisonai_tools.tools (no manual registration). 13 unit
tests cover endpoint selection, result mapping/bounds, and error handling.

Relates to MervinPraison/PraisonAI#5671

Co-authored-by: Mervin Praison <[email protected]>

* test: cover Keenable redirect refusal and bound limits

Address Greptile P2: the protective limits had no focused coverage and
build_opener was fully mocked, so _NoRedirect, the result cap, the
oversized title/url/snippet caps, and the response read limit could be
removed with the suite staying green.

- Lift _NoRedirect to module level so the redirect refusal is importable
  and unit-testable (was a closure inside search()).
- _FakeResponse.read now honours the requested byte count, so the
  oversized-response test exercises the actual read cap.
- Add TestProtections: result-count cap in the request body, title and
  snippet truncation, oversized-URL skip, read-limit enforcement, and
  _NoRedirect dropping a 302 (no follow-up request).
- Add the missing __init__ docstring.

Co-Authored-By: Claude Opus 4.8 <[email protected]>

---------

Co-authored-by: MervinPraison <[email protected]>
Co-authored-by: Mervin Praison <[email protected]>
Co-authored-by: Claude Opus 4.8 <[email protected]>

.github/automation.meta.json

@@ -3,7 +3,7 @@
   "stack": "claude",
   "profile": "tools",
   "mode": "hub",
-  "hubRef": "v1.1.10",
+  "hubRef": "v1.1.11",
   "templateVersion": "1.1.6",
   "installedAt": "2026-09-07T12:54:25Z"
 }

.github/workflows/auto-pr-comment.yml

@@ -27,7 +27,7 @@ jobs:
     permissions:
       issues: write
       pull-requests: write
-    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]0
+    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]1
     secrets: inherit
     with:
-      hub_ref: v1.1.10
+      hub_ref: v1.1.11

.github/workflows/bot-pr-recovery.yml

@@ -16,8 +16,8 @@ permissions:
 
 jobs:
   recover:
-    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]0
+    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]1
     secrets: inherit
     with:
-      hub_ref: v1.1.10
+      hub_ref: v1.1.11
       pr_number: ${{ github.event.inputs.pr_number || '' }}

.github/workflows/ci-failure-claude.yml

@@ -25,8 +25,8 @@ permissions:
 
 jobs:
   detect-and-trigger:
-    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]0
+    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]1
     secrets: inherit
     with:
-      hub_ref: v1.1.10
+      hub_ref: v1.1.11
       pr_number: ${{ github.event.inputs.pr_number || '' }}

.github/workflows/claude-merge-gate.yml

@@ -43,9 +43,9 @@ jobs:
       id-token: write
     steps:
       - uses: actions/checkout@v4
-      - uses: MervinPraison/github-automation-template/.github/actions/[email protected]0
+      - uses: MervinPraison/github-automation-template/.github/actions/[email protected]1
         with:
-          hub_ref: v1.1.10
+          hub_ref: v1.1.11
       - name: Generate GitHub App token
         id: app-token
         uses: actions/create-github-app-token@v1
@@ -87,9 +87,9 @@ jobs:
       issues: write
       actions: read
       id-token: write
-    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]0
+    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]1
     secrets: inherit
     with:
-      hub_ref: v1.1.10
+      hub_ref: v1.1.11
       pr_number: ${{ github.event.inputs.pr_number || github.event.client_payload.pr_number || '' }}
       merge_method: ${{ github.event.inputs.merge_method || 'merge' }}

.github/workflows/claude.yml

@@ -18,7 +18,7 @@ jobs:
       issues: write
       actions: read
       id-token: write
-    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]0
+    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]1
     secrets: inherit
     with:
-      hub_ref: v1.1.10
+      hub_ref: v1.1.11

.github/workflows/merge-conflict-claude.yml

@@ -25,9 +25,9 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-      - uses: MervinPraison/github-automation-template/.github/actions/[email protected]0
+      - uses: MervinPraison/github-automation-template/.github/actions/[email protected]1
         with:
-          hub_ref: v1.1.10
+          hub_ref: v1.1.11
       - name: Detect conflicting PRs and trigger Claude
         uses: actions/github-script@v7
         with:

.github/workflows/pipeline-status-sync.yml

@@ -13,7 +13,7 @@ permissions:
 
 jobs:
   sync:
-    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]0
+    uses: MervinPraison/github-automation-template/.github/workflows/[email protected]1
     secrets: inherit
     with:
-      hub_ref: v1.1.10
+      hub_ref: v1.1.11

praisonai_tools/tools/keenable_search_tool.py

@@ -0,0 +1,175 @@
+"""Keenable Search Tool for PraisonAI Agents.
+
+Keyless (or keyed) web search using Keenable.
+
+Keenable works with no API key and no extra install: it calls the keyless
+public endpoint with the Python standard library (``urllib``). Setting
+``KEENABLE_API_KEY`` switches to the authenticated endpoint for higher limits.
+
+Usage:
+    from praisonai_tools import KeenableSearchTool
+
+    keenable = KeenableSearchTool()
+    results = keenable.search("AI news 2024")
+
+Environment Variables:
+    KEENABLE_API_KEY: Optional. When set, the authenticated endpoint is used.
+"""
+
+import json
+import logging
+import os
+from typing import Any, Dict, List, Optional, Union
+from urllib.request import HTTPRedirectHandler
+
+from praisonai_tools.tools.base import BaseTool
+
+logger = logging.getLogger(__name__)
+
+KEENABLE_SEARCH_URL = "https://api.keenable.ai/v1/search"
+KEENABLE_PUBLIC_SEARCH_URL = "https://api.keenable.ai/v1/search/public"
+KEENABLE_APP_TITLE = "PraisonAI"
+KEENABLE_SEARCH_TIMEOUT_SECONDS = 30
+KEENABLE_MAX_RESPONSE_BYTES = 1_000_000
+KEENABLE_MAX_RESULTS = 50
+KEENABLE_MAX_TITLE_CHARS = 256
+KEENABLE_MAX_URL_CHARS = 2048
+KEENABLE_MAX_SNIPPET_CHARS = 1200
+
+
+def _truncate_text(value: Any, limit: int) -> str:
+    """Trim whitespace and cap ``value`` to ``limit`` characters."""
+    text = str(value or "").strip()
+    if len(text) > limit:
+        return text[:limit - 1] + "\u2026"
+    return text
+
+
+class _NoRedirect(HTTPRedirectHandler):
+    """Redirect handler that refuses to follow any redirect.
+
+    Keenable never redirects. Following a 3xx would resend ``X-API-Key`` to
+    whatever host it points at, so the redirect is dropped (urllib then raises
+    the 3xx as an error, surfaced to the caller) and no second request is sent.
+    """
+
+    def redirect_request(self, *args, **kwargs):
+        """Return ``None`` so urllib does not issue a follow-up request."""
+        return None
+
+
+class KeenableSearchTool(BaseTool):
+    """Tool for keyless (or keyed) web search using Keenable."""
+
+    name = "keenable_search"
+    description = "Search the web using Keenable (keyless; an API key raises limits)."
+
+    def __init__(self, api_key: Optional[str] = None):
+        """Create the tool, resolving the API key from the argument or env.
+
+        Args:
+            api_key: Optional Keenable API key. Falls back to the
+                ``KEENABLE_API_KEY`` environment variable; keyless if unset.
+        """
+        self.api_key = (api_key or os.getenv("KEENABLE_API_KEY") or "").strip()
+        super().__init__()
+
+    def run(
+        self,
+        action: str = "search",
+        query: Optional[str] = None,
+        max_results: int = 5,
+        **kwargs,
+    ) -> Union[str, Dict[str, Any], List[Dict[str, Any]]]:
+        """Dispatch a Keenable action; only ``search`` is supported."""
+        action = action.lower().replace("-", "_")
+        if action == "search":
+            return self.search(query=query, max_results=max_results)
+        return {"error": f"Unknown action: {action}"}
+
+    def search(self, query: str, max_results: int = 5) -> List[Dict[str, Any]]:
+        """Search the web with Keenable.
+
+        Args:
+            query: Search query string.
+            max_results: Maximum results to return (capped at 50).
+
+        Returns:
+            List of dicts with ``title``, ``url``, ``snippet`` and ``provider``,
+            or a single-item list with an ``error`` key on failure.
+        """
+        if not query:
+            return [{"error": "query is required"}]
+
+        result_limit = max(0, min(int(max_results), KEENABLE_MAX_RESULTS))
+        if result_limit == 0:
+            return []
+
+        from urllib.error import URLError
+        from urllib.request import Request, build_opener
+
+        headers = {
+            "Content-Type": "application/json",
+            "Accept": "application/json",
+            "User-Agent": "praisonai-tools",
+            # Required on keyless calls; names the app, carries no user identifier.
+            "X-Keenable-Title": KEENABLE_APP_TITLE,
+        }
+        if self.api_key:
+            headers["X-API-Key"] = self.api_key
+        body = json.dumps({
+            "query": query,
+            "max_results": result_limit,
+            "snippet_max_length": KEENABLE_MAX_SNIPPET_CHARS,
+        }).encode("utf-8")
+        request = Request(
+            KEENABLE_SEARCH_URL if self.api_key else KEENABLE_PUBLIC_SEARCH_URL,
+            data=body,
+            headers=headers,
+            method="POST",
+        )
+
+        try:
+            opener = build_opener(_NoRedirect)
+            with opener.open(request, timeout=KEENABLE_SEARCH_TIMEOUT_SECONDS) as response:
+                raw = response.read(KEENABLE_MAX_RESPONSE_BYTES + 1)
+        except URLError as exc:
+            logger.error(f"Keenable search error: {exc}")
+            return [{"error": str(exc)}]
+        except Exception as exc:
+            logger.error(f"Keenable search error: {exc}")
+            return [{"error": str(exc)}]
+
+        if len(raw) > KEENABLE_MAX_RESPONSE_BYTES:
+            return [{"error": "Keenable returned an oversized response"}]
+
+        try:
+            payload = json.loads(raw)
+        except ValueError as exc:
+            return [{"error": f"Keenable returned invalid JSON: {exc}"}]
+        if not isinstance(payload, dict) or not isinstance(payload.get("results"), list):
+            return [{"error": "Keenable returned an invalid response shape"}]
+
+        results = []
+        for result in payload["results"]:
+            if not isinstance(result, dict):
+                continue
+            url = result.get("url")
+            if not isinstance(url, str) or not url or len(url) > KEENABLE_MAX_URL_CHARS:
+                continue
+            # The page text is in `snippet`; `description` is usually empty.
+            snippet = " ".join(str(result.get("snippet") or result.get("description") or "").split())
+            results.append({
+                "title": _truncate_text(result.get("title", ""), KEENABLE_MAX_TITLE_CHARS),
+                "url": url,
+                "snippet": _truncate_text(snippet, KEENABLE_MAX_SNIPPET_CHARS),
+                "provider": "keenable",
+            })
+            if len(results) >= result_limit:
+                break
+        return results
+
+
+def keenable_search(query: str, max_results: int = 5) -> List[Dict[str, Any]]:
+    """Search the web with Keenable (keyless; KEENABLE_API_KEY raises limits)."""
+    return KeenableSearchTool().search(query=query, max_results=max_results)

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "praisonai-tools"
-version = "0.2.44"
+version = "0.2.45"
 description = "Extended tools for PraisonAI Agents"
 authors = [
     {name = "Mervin Praison"}

tests/unit/tools/test_keenable_search_tool.py

@@ -0,0 +1,273 @@
+"""Unit tests for KeenableSearchTool."""
+
+import json
+import os
+from unittest.mock import MagicMock, patch
+
+from praisonai_tools.tools.keenable_search_tool import (
+    KEENABLE_MAX_RESULTS,
+    KEENABLE_MAX_SNIPPET_CHARS,
+    KEENABLE_MAX_TITLE_CHARS,
+    KEENABLE_MAX_URL_CHARS,
+    KEENABLE_PUBLIC_SEARCH_URL,
+    KEENABLE_SEARCH_URL,
+    KeenableSearchTool,
+    _NoRedirect,
+    keenable_search,
+)
+
+
+class _FakeResponse:
+    """Minimal stand-in for an ``http.client.HTTPResponse`` context manager."""
+
+    def __init__(self, raw: bytes):
+        self._raw = raw
+
+    def read(self, amt=None):
+        """Return at most ``amt`` bytes, honouring the caller's read cap."""
+        if amt is None:
+            return self._raw
+        return self._raw[:amt]
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *_args):
+        return False
+
+
+def _payload(results):
+    return json.dumps({"results": results}).encode("utf-8")
+
+
+# ── Import + schema (CI smoke) ──────────────────────────────────────
+
+
+class TestImportAndSchema:
+    def test_import_via_package(self):
+        from praisonai_tools import KeenableSearchTool as PkgTool
+
+        assert PkgTool is KeenableSearchTool
+
+    def test_schema_generation(self):
+        tool = KeenableSearchTool()
+        schema = tool.get_schema()
+        assert schema["function"]["name"] == "keenable_search"
+        assert "description" in schema["function"]
+        assert "action" in schema["function"]["parameters"]["properties"]
+
+
+# ── Endpoint selection ──────────────────────────────────────────────
+
+
+class TestEndpointSelection:
+    def test_public_endpoint_without_key(self):
+        with patch.dict(os.environ, {}, clear=True):
+            tool = KeenableSearchTool()
+        assert tool.api_key == ""
+
+        captured = {}
+
+        def fake_open(request, timeout=None):
+            captured["url"] = request.full_url
+            captured["headers"] = dict(request.header_items())
+            return _FakeResponse(_payload([]))
+
+        opener = MagicMock()
+        opener.open.side_effect = fake_open
+        with patch("urllib.request.build_opener", return_value=opener):
+            tool.search("python", max_results=3)
+
+        assert captured["url"] == KEENABLE_PUBLIC_SEARCH_URL
+        header_names = {k.lower() for k in captured["headers"]}
+        assert "x-api-key" not in header_names
+        assert "X-keenable-title" in captured["headers"] or "X-Keenable-Title" in captured["headers"]
+
+    def test_authenticated_endpoint_with_key(self):
+        tool = KeenableSearchTool(api_key="secret-key")
+        captured = {}
+
+        def fake_open(request, timeout=None):
+            captured["url"] = request.full_url
+            captured["headers"] = dict(request.header_items())
+            return _FakeResponse(_payload([]))
+
+        opener = MagicMock()
+        opener.open.side_effect = fake_open
+        with patch("urllib.request.build_opener", return_value=opener):
+            tool.search("python", max_results=3)
+
+        assert captured["url"] == KEENABLE_SEARCH_URL
+        # urllib title-cases header keys.
+        assert captured["headers"].get("X-api-key") == "secret-key"
+
+
+# ── Result mapping + bounds ─────────────────────────────────────────
+
+
+class TestResultMapping:
+    def _run(self, tool, results):
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(_payload(results))
+        with patch("urllib.request.build_opener", return_value=opener):
+            return tool.search("q", max_results=10)
+
+    def test_maps_fields(self):
+        out = self._run(
+            KeenableSearchTool(),
+            [{"title": "T", "url": "https://example.com", "snippet": "hello world"}],
+        )
+        assert out == [{
+            "title": "T",
+            "url": "https://example.com",
+            "snippet": "hello world",
+            "provider": "keenable",
+        }]
+
+    def test_skips_entries_without_url(self):
+        out = self._run(
+            KeenableSearchTool(),
+            [{"title": "no url"}, {"title": "ok", "url": "https://a.com"}],
+        )
+        assert len(out) == 1
+        assert out[0]["url"] == "https://a.com"
+
+    def test_respects_max_results(self):
+        tool = KeenableSearchTool()
+        many = [{"url": f"https://a.com/{i}"} for i in range(10)]
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(_payload(many))
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = tool.search("q", max_results=3)
+        assert len(out) == 3
+
+    def test_zero_max_results_short_circuits(self):
+        out = KeenableSearchTool().search("q", max_results=0)
+        assert out == []
+
+    def test_empty_query_errors(self):
+        out = KeenableSearchTool().search("", max_results=5)
+        assert out == [{"error": "query is required"}]
+
+
+# ── Error handling ──────────────────────────────────────────────────
+
+
+class TestErrorHandling:
+    def test_invalid_shape_errors(self):
+        tool = KeenableSearchTool()
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(b'{"not_results": 1}')
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = tool.search("q")
+        assert out and "error" in out[0]
+
+    def test_oversized_response_errors(self):
+        tool = KeenableSearchTool()
+        oversized = b"x" * (1_000_001 + 1)
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(oversized)
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = tool.search("q")
+        assert out == [{"error": "Keenable returned an oversized response"}]
+
+    def test_http_error_is_captured(self):
+        tool = KeenableSearchTool()
+        opener = MagicMock()
+        opener.open.side_effect = RuntimeError("boom")
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = tool.search("q")
+        assert out == [{"error": "boom"}]
+
+
+# ── Protective bounds + redirect refusal ────────────────────────────
+
+
+class TestProtections:
+    def _capture_body(self, tool, requested):
+        """Run a search and return the decoded JSON request body sent."""
+        captured = {}
+
+        def fake_open(request, timeout=None):
+            captured["body"] = json.loads(request.data.decode("utf-8"))
+            return _FakeResponse(_payload([]))
+
+        opener = MagicMock()
+        opener.open.side_effect = fake_open
+        with patch("urllib.request.build_opener", return_value=opener):
+            tool.search("q", max_results=requested)
+        return captured["body"]
+
+    def test_caps_requested_results_at_limit(self):
+        """max_results above the hard cap is clamped before the request."""
+        body = self._capture_body(KeenableSearchTool(), requested=1000)
+        assert body["max_results"] == KEENABLE_MAX_RESULTS
+
+    def test_truncates_oversized_title(self):
+        """Titles longer than the cap are truncated with an ellipsis."""
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(
+            _payload([{"title": "T" * 5000, "url": "https://a.com"}])
+        )
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = KeenableSearchTool().search("q")
+        assert len(out[0]["title"]) == KEENABLE_MAX_TITLE_CHARS
+        assert out[0]["title"].endswith("…")
+
+    def test_truncates_oversized_snippet(self):
+        """Snippets longer than the cap are truncated with an ellipsis."""
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(
+            _payload([{"snippet": "s " * 5000, "url": "https://a.com"}])
+        )
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = KeenableSearchTool().search("q")
+        assert len(out[0]["snippet"]) == KEENABLE_MAX_SNIPPET_CHARS
+        assert out[0]["snippet"].endswith("…")
+
+    def test_skips_oversized_url(self):
+        """Results whose URL exceeds the length cap are dropped."""
+        long_url = "https://a.com/" + "x" * KEENABLE_MAX_URL_CHARS
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(
+            _payload([{"url": long_url}, {"url": "https://ok.com"}])
+        )
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = KeenableSearchTool().search("q")
+        assert out == [{
+            "title": "",
+            "url": "https://ok.com",
+            "snippet": "",
+            "provider": "keenable",
+        }]
+
+    def test_read_is_capped_at_limit(self):
+        """The body read is bounded, so an oversized stream still errors out."""
+        tool = KeenableSearchTool()
+        # Far larger than the cap; _FakeResponse.read honours the byte count,
+        # so this verifies the read limit rather than a pre-sized buffer.
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(b"x" * 5_000_000)
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = tool.search("q")
+        assert out == [{"error": "Keenable returned an oversized response"}]
+
+    def test_refuses_redirects(self):
+        """The redirect handler drops 3xx responses so no follow-up is sent."""
+        handler = _NoRedirect()
+        # Returning None is urllib's signal to NOT build/send a second request.
+        assert handler.redirect_request(
+            MagicMock(), MagicMock(), 302, "Found", {}, "https://evil.example"
+        ) is None
+
+
+# ── Module-level convenience function ───────────────────────────────
+
+
+class TestConvenienceFunction:
+    def test_keenable_search_function(self):
+        opener = MagicMock()
+        opener.open.return_value = _FakeResponse(_payload([{"url": "https://a.com"}]))
+        with patch("urllib.request.build_opener", return_value=opener):
+            out = keenable_search("q", max_results=2)
+        assert out[0]["provider"] == "keenable"