fix: background task media delivery + vision download timeout

[teknium1]

Summary

Two bug fixes salvaged from PR #2269 by amethystani. The third fix (opencode-go auth) was already resolved on main.

Fix 1: Background task media delivery (gateway/run.py)

adapter.send_file() → adapter.send_document(). The send_file method doesn't exist on BasePlatformAdapter — the call raised AttributeError which was silently swallowed by except Exception: pass. Background task media files were never delivered to users.

Fix 2: Vision image download timeout (tools/vision_tools.py)

• Configurable via HERMES_VISION_DOWNLOAD_TIMEOUT env var (default 30s). Separate from HERMES_VISION_TIMEOUT which governs the LLM API call.
• Guard against raise None when max_retries=0 — now raises a descriptive RuntimeError instead of TypeError.

Verification

• 81 vision + background task tests pass
• Both files compile clean

Closes #2269. Credit to amethystani.

[github-actions]

⚠️ Supply Chain Risk Detected

This PR contains patterns commonly associated with supply chain attacks. This does not mean the PR is malicious — but these patterns require careful human review before merging.

⚠️ WARNING: base64 encoding/decoding detected

Base64 has legitimate uses (images, JWT, etc.) but is also commonly used to obfuscate malicious payloads. Verify the usage is appropriate.

Matches (first 20):

2393:+                b64 = base64.b64encode(content.encode("utf-8")).decode("ascii")

⚠️ WARNING: exec() or eval() usage

Dynamic code execution can hide malicious behavior, especially when combined with base64 or network fetches.

Matches (first 20):

2179:+                home = self._sandbox.process.exec("echo $HOME").result.strip()
2338:+        content into the sandbox via exec(), so new/updated credentials are

---

Automated scan triggered by supply-chain-audit. If this is a false positive, a maintainer can approve after manual review.

[github-actions]

⚠️ Supply Chain Risk Detected

This PR contains patterns commonly associated with supply chain attacks. This does not mean the PR is malicious — but these patterns require careful human review before merging.

⚠️ WARNING: base64 encoding/decoding detected

Base64 has legitimate uses (images, JWT, etc.) but is also commonly used to obfuscate malicious payloads. Verify the usage is appropriate.

Matches (first 20):

2407:+                b64 = base64.b64encode(content.encode("utf-8")).decode("ascii")

⚠️ WARNING: exec() or eval() usage

Dynamic code execution can hide malicious behavior, especially when combined with base64 or network fetches.

Matches (first 20):

2193:+                home = self._sandbox.process.exec("echo $HOME").result.strip()
2352:+        content into the sandbox via exec(), so new/updated credentials are

---

Automated scan triggered by supply-chain-audit. If this is a false positive, a maintainer can approve after manual review.