Bucket B — Draft close-comments for your review

Nothing below is posted yet. Review, then tell me "post all", "post all except #X, #Y", or edit any wording. I'll then use gh issue comment + gh issue close for each.

Legend: 🟢 = safe to close · 🟡 = borderline, recommend a look · ⚪ = recommend KEEP OPEN (real request buried in promo)

---

1) Stale / already-fixed (10) — all verified on current master

Template: confirm the fix + invite reopen. Then close as completed.

🟢 #1079 — LLM retry loop returned on first attempt

This has been resolved — the LLM provider layer (gpt_researcher/utils/llm.py / llm_provider/generic/base.py) was rewritten with a proper retry/attempt loop and exception handling. Closing as fixed; please reopen if you still hit this on the latest version. Thanks for the report!

🟢 #1098 — o3 / reasoning models reject temperature

Reasoning models are now handled via NO_SUPPORT_TEMPERATURE_MODELS in gpt_researcher/llm_provider/generic/base.py, which strips the unsupported temperature arg for o1/o3/o4/gpt-5. Closing as fixed — note that for LiteLLM you'll need the exact model name to match. Reopen if it persists.

🟢 #1105 — Inconsistent required Python versions

Resolved — pyproject.toml now declares requires-python = ">=3.11", matching the README. Closing.

🟢 #1131 — Long prompts → FileNotFoundError (filename too long)

Fixed — sanitize_filename() in backend/server/server_utils.py now hashes the task portion (md5[:10]) instead of embedding the full prompt, so filenames stay short. Closing as fixed.

🟡 #1232 — npm package doesn't support https/wss host

A contributor noted https/wss support was since added to the npm package. Closing as resolved — please reopen if you still see ws:// being forced on an https host. (Note: I verified the Python side but not the published npm package version — wording kept soft.)

🟢 #1254 — parse_dimension incorrect on decimals

Fixed — parse_dimension() in gpt_researcher/scraper/utils.py now does int(float(value)), handling decimal values like 409.12. (Percent values are still out of scope.) Closing.

🟢 #1279 — 'list' object has no attribute 'split' in count_words

Fixed — count_words() in gpt_researcher/skills/deep_research.py now detects list input and joins it before counting. Closing as fixed; reopen if you reproduce on the latest version.

🟢 #1391 — Could not connect to MCP Server via HTTP

The MCP server moved to its own repo: https://github.com/assafelovic/gptr-mcp. Closing here — please file MCP-server issues in that repo. Thanks!

🟢 #1398 — Runtime error (temperature None for ChatOpenAI)

Same root cause as #1098 — reasoning models now strip temperature via NO_SUPPORT_TEMPERATURE_MODELS in base.py. Closing as fixed; reopen if it persists on the latest version.

🟢 #1415 — 'PromptFamily' object has no attribute 'curate_soures' (typo)

Fixed — the typo is gone; the codebase consistently uses curate_sources now. Closing as resolved.

---

2) Duplicates (5)

🟢 #1685 — anybrowse MCP integration

Duplicate of #1684 — consolidating there. Closing.

🟢 #1707 — Security audit findings (tool description injection)

Duplicate of #1706 — same report. Closing to keep it in one place.

🟢 #1744 — Merxex Integration: Agent-to-Agent Commerce

Duplicate of #1736 — closing, let's keep one thread.

🟢 #1745 — Proposal: Agent-to-Agent Commerce via Merxex

Duplicate of #1736 — closing.

🟢 #1747 — Standardizing Agent Commerce: Merxex Proposal

Duplicate of #1736 — closing.

---

3) Promotional / out-of-scope (22)

Template (polite decline): "Thanks for reaching out. This reads as a promotion for an external product/service rather than a bug or a concrete, open feature request, so I'm closing it to keep the tracker focused. If there's a specific, open-source integration you'd like to propose, a PR or a Discussions thread is the best path. 🙏"

🟢 Close as promo (apply template): #1629 (MoltBridge), #1687 (WheatCoin), #1689 (AgentID), #1699 (AI Village), #1702 (ResearchClawBench — or redirect to Discussions), #1706 (paid "$29 audit"), #1711 (AgentFolio), #1714 (Clarvia badge), #1716 (Clarvia MCP), #1728 (veritasacta receipts), #1729 (Loaditout badge), #1732 (trust-score promos), #1733 (AgentWeb), #1736 (Merxex), #1789 (HVTracker), #1796 (Agent Magnet), #1801 (USDC federation), #1808 (BGPT)

Borderline — recommend a quick look before closing:

⚪ #1076 — Add open-source provider AI/ML API — RECOMMEND KEEP OPEN. This is a real, generic LLM-provider request (aimlapi.com). GPTR already supports many providers via LiteLLM; could be a clean community PR. Suggest labeling good-first-issue/provider and inviting a PR rather than closing.

🟡 #1684 — anybrowse for Cloudflare-protected scraping — The product pitch is promo, but the underlying bug is real: scrapers silently return on Cloudflare 403s. Suggested reply:

The specific tool is out of scope, but the underlying problem — scrapers silently failing on Cloudflare-protected (403) pages — is valid. I'll repurpose this as a tracked bug for a generic fallback/clear-error path. (Then relabel as bug instead of closing.)

🟡 #1704 — Plasmate extraction / 🟡 #1751 — SerpBase SERP — Vendor self-promos (SerpBase had a sockpuppet comment). Recommend close as promo unless you want generic "cheaper scraper/SERP option" tracked. Default: close.

---

Proposed actions summary

• Close now (32): all 10 stale + 5 duplicates + 17 clear promo (#1702 your call on redirect-to-Discussions).
• Keep open / relabel (3): #1076 (provider PR invite), #1684 (convert to bug), + your call on #1704/#1751.