PluginDaemonInternalServerError: failed to get plugin runtime when invoking tool with file upload via remote debug

[pareek-ml]

Environment

• Remote debug install (debug.dify.ai:5003)
• dify-plugin SDK v0.9.1
• Python 3.12

Steps to reproduce
A router workflow calls a tool with command="add" and an uploaded .zip file (passed via the files param). The plugin process is confirmed running locally (python -m main logs Installed tool: ...).

Error
PluginDaemonInternalServerError: failed to get plugin runtime — occurring before any of the tool's own code executes.

What's been ruled out

• manifest.yaml validated against the current SDK's Pydantic models (PluginConfiguration, I18nObject, etc.) — fully compliant
• Simulated the exact file-download + zip-extraction flow locally using the real upload URL — succeeds without error
• Added debug instrumentation directly in the tool's _invoke to confirm request routing/reachability

manifest.yaml

version: 0.0.2
type: plugin
author: pareek-ml
name: agent_skills
label:
  en_US: agent_skills
description:
  en_US: allows you to install and use custom skills
icon: icon.svg
icon_dark: icon-dark.svg
resource:
  memory: 268435456
  permission:
    tool:
      enabled: true
    model:
      enabled: true
      llm: true
      text_embedding: false
      rerank: false
      tts: false
      speech2text: false
      moderation: false
    endpoint:
      enabled: true
    app:
      enabled: true
    storage:
      enabled: true
      size: 104857600
plugins:
  tools:
    - provider/agent_skills.yaml
meta:
  version: 0.0.2
  arch:
    - amd64
    - arm64
  runner:
    language: python
    version: "3.12"
    entrypoint: main
  minimum_dify_version: null
created_at: 2026-07-15T16:56:51.72279+02:00
privacy: PRIVACY.md
verified: false

[dosubot]

Hi @pareek-ml! I'm Dosu and I'm helping the Dify team.

The failed to get plugin runtime error originates in createSession — it fires when manager.GetPluginRuntime() can't find your plugin in the debuggingPluginRuntime map. Unfortunately the error is generic and swallows the underlying cause, which makes it hard to diagnose.

For remote debug plugins, the runtime lookup works by checking the debuggingPluginRuntime map first, then localPluginRuntimes [1]. Several things can prevent your plugin from being in that map even though it's running locally:

1. Plugin identity mismatch — A significant fix landed in PR #731 (May 2026) that unified the plugin_id format between remote debug and local plugins. Before that fix, RemotePluginRuntime.Identity() could generate a different identifier than what the lookup expected. Make sure you're running a daemon version that includes this fix.

2. Handshake timeout — Remote plugins must complete the full registration handshake (manifest, tool declarations, assets, END event) within 10 seconds of connecting, or the connection is dropped [2]. If your plugin is large or the network is slow, it may silently fail registration.

3. Cluster sync / cache staleness — PR #589 made RegisterPlugin idempotent to handle reconnections, and PR #667 fixed stale cache entries persisting after disconnect. If you're on an older daemon, reconnecting after a drop could leave the registry in an inconsistent state.

4. Redirect logic — PR #761 (June 2026) fixed remote debug redirect behavior so that when a runtime isn't found locally it correctly redirects rather than erroring.

What to check:

• Confirm the daemon version you're connecting to (debug.dify.ai:5003) includes the fixes above (post-June 2026).
• Watch the daemon-side logs during plugin connection — look for whether the handshake completes successfully and the runtime is stored.
• Verify that the unique_identifier the daemon resolves for your plugin matches what's being queried at invocation time (the identity is derived from author/name:version in the manifest).

The file upload aspect is likely a red herring — the error fires before your tool code runs because the session can't even locate the runtime. The root issue is almost certainly a registration/identity problem on the daemon side.

To reply, just mention @dosu.

---

Docs are dead. Just use Dosu.

[fatelei]

how about your dify-plugin-daemon version? and please submit your difypkg file