Source-backed Japanese politics evidence from five national newspapers, ready for Dify summaries.
Source: Dify plugin source.
Support: [email protected].
Retrieve source-backed evidence for Japanese politics through AIsa, then use a Dify
LLM or Agent to write a Japanese digest with citations. This independently
installable plugin contains one provider and one tool, both named .
Japanese · Privacy
Install the packaged in Dify Integrations, authorize the provider with
your AIsa API key, and add the tool to a workflow or Agent. Credential
validation is local: it checks the key format without an API request and does not
verify account balance or remote authorization. Requests use the fixed AIsa API
endpoint ; publisher credentials and subscription cookies are
not accepted.
The tool searches Yomiuri, Asahi, Nikkei, Mainichi, and Sankei. The default makes a separate search lane for each newspaper. Source coverage is reported individually; a story covered by multiple newspapers can retain those citations without implying agreement. Selecting one newspaper narrows the requested coverage.
Tavily news search is the primary retrieval route through AIsa. Enabling Brave
fallback or extraction can add billable API requests. Use to
inspect the plan first. The tool has bounded retrieval; it cannot guarantee every
article from a newspaper or immediate indexing of breaking news.
The tool returns JSON evidence plus a readable Japanese source overview, not a
model-generated narrative. Connect the following named outputs to downstream
nodes: , , , , and . The full JSON
also records the time window, request activity, and request plan.
reports each selected lane with , , , and
error information. The overall status is , , , , or
. An empty or inaccessible source is reported explicitly. A request
failure must not be described as an absence of news.
Suggested workflow: tool → LLM → answer/output. Pass , , and
into the model prompt, together with this instruction:
Write a concise Japanese news digest using only the supplied evidence. Cite the
source URL for every item. Distinguish facts from opinion. Do not infer missing
details from a headline or excerpt. State missing source coverage and unknown
publication dates. Do not claim a paywalled article's full text was read.
The plugin does not invoke a model or request Dify model, app, endpoint, or storage
permissions. Model choice and any model charges belong to the downstream workflow.
Search snippets, public excerpts, and extracted content have different evidential
limits. Subscriber-only content, robots restrictions, timeouts, missing timestamps,
and indexing delays can produce partial results. Article URLs remain linked to
the publisher; extraction does not bypass a subscription or retrieve content with
publisher login credentials. Unknown publication dates are disclosed rather than
presented as verified fresh stories. Duplicate URLs are removed before output.
From this directory, create a package-local Python 3.12 environment and
install dependencies:
From the repository root, run the package's mocked tests and collection check:
The tests use fixed clocks and fake clients. They do not need a real key or send
paid API requests. Keep , virtual environments, caches, and built archives
out of Git.
For remote Dify debugging, copy to , replace its placeholders
with the debug host and key from your workspace, and run inside
this directory. A connected debug plugin can be invoked by other workspace users;
news retrieval then uses your configured AIsa credentials. Do not share the debug
key. End-to-end validation requires an authorized workspace and live API usage.
From the repository root:
The native identity is and the plugin version is .
Packaging creates an installable archive; publishing, repository pushes, and
Marketplace submission are separate actions.
Some outbound calls build their address at runtime, so the list may be incomplete.
Last checked 10/29/2026 07:42 AM · version 0.1.0