feat: add Keenable as a configurable internet search backend

[ilya-bogin-keenable]

Summary

Adds Keenable as a new internet search backend alongside the existing Bocha / Tavily / Google / Bing / Xinyu retrievers, following the Tavily backend pattern (#1357). Additive and opt-in via INTERNET_SEARCH_BACKEND=keenable; existing backends are untouched.

Keenable is a web search API built for AI agents. Unlike the key-required backends it is keyless by default: with no key it calls the public endpoint (rate-limited), and an optional KEENABLE_API_KEY only lifts the cap.

Files changed

• src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (new): InternetKeenableRetriever. No SDK dependency, a thin requests call. Keyless requests hit /v1/search/public; a configured key switches to /v1/search with an X-API-Key header. Attribution via X-Keenable-Title. Results map into TextualMemoryItem the same way the Tavily retriever does.
• src/memos/configs/internet_retriever.py: KeenableSearchConfig (API key optional) + registration in InternetRetrieverConfigFactory.
• src/memos/memories/textual/tree_text_memory/retrieve/internet_retriever_factory.py: register keenable + constructor branch.
• src/memos/api/config.py: INTERNET_SEARCH_BACKEND=keenable branch (KEENABLE_API_KEY optional).

Testing

• python -m py_compile on all changed files: passes.
• ruff check on all changed files: passes.

[ilya-bogin-keenable]

Hey @bittergreen could you please take a look? Thanks!

[Memtensor-AI]

🤖 Open Code Review

Target: PR #2072
Task: 1a3c8d0e9d69e77a
Base: main
Head: feat/keenable-web-search

🔍 OpenCodeReview found 7 issue(s) in this PR.

---

1. src/memos/configs/internet_retriever.py (L90-L92)

The base class BaseInternetRetrieverConfig declares api_key: str as a required field (Field(...)). This override changes it to str | None = Field(default=None), which is intentional for keyless support — but config.py passes api_key: os.getenv("KEENABLE_API_KEY", ""), supplying an empty string "" instead of None when the env var is unset.

In keenablesearch.py, the retriever normalizes this with self.api_key = (api_key or "").strip(), so an empty string ends up selecting the public (unauthenticated) path correctly. However, Pydantic will accept "" as a valid str, not coerce it to None, so the field value will be "" rather than None after construction. This is harmless at runtime (the retriever handles it), but the config object's api_key will be "" rather than None in the keyless case, which is inconsistent with what the field type (str | None) and its description imply.

Consider either:

1. Using default=None consistently and passing os.getenv("KEENABLE_API_KEY") or None in config.py, or
2. Adding a Pydantic field validator on KeenableSearchConfig to coerce empty strings to None.

---

2. src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (L60)

A single TextRank instance is shared across all 8 worker threads in ContextThreadPoolExecutor. TextRank.textrank() builds internal graph structures and updates mutable instance state during each call, which is not thread-safe. Concurrent calls will race on that shared state, producing corrupted keyword results or intermittent crashes.

Fix: instantiate a fresh TextRank inside _process_result (one per call) or protect the shared instance with a threading.Lock.

# Option A – per-call instantiation (simplest, avoids lock)
from jieba.analyse import TextRank
tags = TextRank().textrank(summary, topK=3)[:3]

# Option B – lock around the shared instance
import threading
self._textrank_lock = threading.Lock()
# … in _process_result:
with self._textrank_lock:
    tags = self.zh_fast_keywords_extractor.textrank(summary, topK=3)[:3]

---

3. src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (L297)

self.embedder.embed() is called concurrently from up to 8 threads. OllamaEmbedder holds a single ollama.Client instance (self.client) and mutates self.config during __init__. The ollama.Client uses an httpx session internally; concurrent embed() calls on the same client instance may race on connection-pool state.

Beyond correctness, silent failure is the bigger concern: _process_result only logs the exception, so garbled embeddings would be silently stored as valid memories.

Fix: confirm ollama.Client is thread-safe for concurrent embed() calls (it uses httpx.Client which is documented thread-safe for requests), or serialize access with a lock. At minimum, surface embedding failures clearly rather than swallowing them.

---

4. src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (L182-L199)

The try block is too broad: it catches ValueError from malformed JSON, KeyError from an unexpected response shape, and any programming mistake inside raw_results processing — conflating logic bugs with genuine network errors. All failures silently return [], making root cause diagnosis difficult.

Also, import traceback inside the except block should be a module-level import.

Suggested narrowing:

import traceback  # move to top of file

try:
    resp = requests.post(..., timeout=self.timeout)
    resp.raise_for_status()
except requests.RequestException:
    logger.error("Keenable search error: %s", traceback.format_exc())
    return []

raw_results = resp.json().get("results", [])[:limit]

---

5. src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (L71-L73)

The keywords dict is a large constant literal reconstructed on every _extract_tags call. Since _process_result calls _extract_tags concurrently across 8 threads, this dict is repeatedly re-allocated in a hot path with no benefit.

Promote it to a module-level constant:

_CATEGORY_KEYWORDS: dict[str, list[str]] = {
    "economy": ["economy", "gdp", ...],
    ...
}

Then reference _CATEGORY_KEYWORDS inside _extract_tags.

---

6. src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (L198)

Eagerly evaluated f-string passed to logger.error: traceback.format_exc() is called unconditionally even when the error log level is disabled. Use %-formatting to defer evaluation:

logger.error("Keenable search error: %s", traceback.format_exc())

---

7. src/memos/memories/textual/tree_text_memory/retrieve/keenablesearch.py (L235)

Same eagerly evaluated f-string pattern. Prefer:

logger.error("Error processing Keenable search result: %s", e)

Generated by cloud-assistant via Open Code Review.

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-33586be3c2b3f89c-20260801031509: 107/107 passed — these do NOT affect the PR verdict; review the branch manually.

Branch: feat/keenable-web-search

[ilya-bogin-keenable]

Аixed 4 of the 5: jieba is now gated behind require_python_package like the bocha retriever (an English-only install used to fail at construction), empty content is no longer embedded, and GDP/AI are lowercased so they can actually match.

Will not fix for #2. Forwarding the local mode to the API would break the default path: the search endpoint returns 400 for mode: "fast", and fast is the default the caller passes. The API's mode and this method's mode are different things, so I added a comment saying so instead.

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-b3c50e54d556e024-20260804132254: 86/87 passed, 1 failed — these do NOT affect the PR verdict; review the branch manually.

Branch: feat/keenable-web-search

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-b90c9e651259700e-20260813203716: 140/140 passed — these do NOT affect the PR verdict; review the branch manually.

Branch: feat/keenable-web-search

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-504091e23c505b41-20260814164055: 173/174 passed, 1 failed — these do NOT affect the PR verdict; review the branch manually.

Branch: feat/keenable-web-search

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 5s

Branch: feat/keenable-web-search

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-cebeff6b3ed746cf-20260818124755: 139/140 passed — these do NOT affect the PR verdict; review the branch manually.

Branch: feat/keenable-web-search

[Memtensor-AI]

✅ Automated Test Results: PASSED

All tests passed (5/5 executed). memos_python_core/changed-python-source: 5/5. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-d929796b8f2bd83f-20260819141815: 136/136 passed — these do NOT affect the PR verdict; review the branch manually.

Branch: feat/keenable-web-search

[Memtensor-AI]

⚠️ Automated Test Results: ENV ISSUE

The test environment encountered an issue that requires manual attention.

Details: Executor error: Command failed: git clone --depth 1 --branch feat/keenable-web-search git@github.com:keenableai/MemOS.git /data/test-workspaces/0bb2eb3c8566c1ed/repo
Cloning into '/data/test-workspaces/0bb2eb3c8566c1ed/repo'...
Branch: feat/keenable-web-search

[Memtensor-AI]

⚠️ Automated Test Results: ENV ISSUE

The test environment encountered an issue that requires manual attention.

Details: Executor error: Command failed: git clone --depth 1 --branch feat/keenable-web-search git@github.com:keenableai/MemOS.git /data/test-workspaces/1a3c8d0e9d69e77a/repo
Cloning into '/data/test-workspaces/1a3c8d0e9d69e77a/repo'...
kex_exchange_identification: Connection closed by remote host
Connection closed by UNKNOWN port 65535
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.
Branch: feat/keenable-web-search