flemming-it
3422fe526d
All checks were successful
CI / Linux x86_64 (Forgejo) (push) Successful in 1m32s
Ollama-backed summarisation module — fourth real F∆I capability,
following text.extract / llm.chat / text.translate. Completes
the text-processing trio for compose-grade workflows
(extract -> translate -> summarise, or any subset).
Capability:
Inputs:
text : text (source text)
style : text (e.g. "one paragraph", "three bullets";
defaults to "one paragraph")
language : text (output language hint; empty = source)
endpoint : text (Ollama /api/chat URL)
model : text
api_key : text (optional)
Outputs:
summary : text
style : text (echo for audit)
language : text (echo for audit)
model_endpoint : text
model_name : text
model_digest : text (Ollama /api/show probe; empty for
non-Ollama or transient failures)
Permissions: net to localhost / 127.0.0.1 / api.openai.com /
api.anthropic.com.
System prompt is conservative: faithful, neutral, preserves
named entities, no editorialising, no metadata. Reuses the LLM
client + digest-probe pattern from llm-chat / text-translate;
the duplication is acceptable at four modules and will be
factored into a shared crate when a fifth lands.
15 host-side tests cover prompt-building (style + language +
empty-language), Ollama-shaped response parsing, URL transform,
digest extraction, end-to-end success / probe-failure / non-
Ollama paths.
Wasm artifact: 118 KB. Built with v1.0 fai:platform imports.
Bootstrapped via 'fai new module text.summarize'.
Signed-off-by: flemming-it <sf@flemming.it>
48 lines
1.2 KiB
YAML
48 lines
1.2 KiB
YAML
schema_version: 1
|
|
name: text-summarize
|
|
version: 0.1.0
|
|
|
|
# Capability provided by this module.
|
|
provides:
|
|
- capability: text.summarize
|
|
version: 0.1.0
|
|
|
|
# Inputs the invoke function accepts.
|
|
inputs:
|
|
# Source text to summarise.
|
|
text: text
|
|
# Style / shape of the summary, e.g. "one paragraph",
|
|
# "three bullet points", "a tweet". Defaults to
|
|
# "one paragraph" when omitted.
|
|
style: text
|
|
# Optional output language hint, e.g. "German", "ja-JP".
|
|
# Empty = same language as the source.
|
|
language: text
|
|
# Ollama-shaped /api/chat endpoint.
|
|
endpoint: text
|
|
# Model identifier (e.g. "qwen2.5:14b").
|
|
model: text
|
|
# Optional bearer token for cloud-hosted endpoints.
|
|
api_key: text
|
|
|
|
# Outputs produced.
|
|
outputs:
|
|
# The summary text.
|
|
summary: text
|
|
# Echo of the style input for downstream audit.
|
|
style: text
|
|
# Echo of the language input.
|
|
language: text
|
|
# Endpoint the summary was generated against.
|
|
model_endpoint: text
|
|
# Model identifier as supplied to the LLM API.
|
|
model_name: text
|
|
# SHA-256 digest of the served Ollama model (when reachable).
|
|
model_digest: text
|
|
|
|
# Permissions required.
|
|
permissions:
|
|
- "net: localhost"
|
|
- "net: 127.0.0.1"
|
|
- "net: api.openai.com"
|
|
- "net: api.anthropic.com"
|