API Reference

API Reference

All exports are available from the main entry point:

import { ... } from 'agentic-rss-parser';

---

createParser(options?)

Returns a ParserCompat instance — drop-in replacement for rss-parser.

createParser(options?: ParserOptions): Parser

ParserOptions

Field	Type	Default	Description
normalize	boolean	true	Normalize field names
customFields.feed	Array	[]	Extra feed-level fields to extract
customFields.item	Array	[]	Extra item-level fields to extract
headers	Record<string,string>	—	Extra HTTP request headers
timeout	number	10000	HTTP timeout in ms
maxRedirects	number	5	Max HTTP redirects to follow

---

Parser instance methods

parser.parseFeed(urls, config?)

Full agentic pipeline — fetch, parse, deduplicate, analyse.

parseFeed(urls: string | string[], config?: ParseFeedConfig): Promise<FeedResultsArray>

ParseFeedConfig

Field	Type	Default	Description
dbPath	string	cwd/data/rss-agent.db	SQLite DB path
storage	StorageAdapter	—	Custom adapter (overrides dbPath)
fetchFullArticle	boolean	false	Fetch + strip full article HTML
concurrency	number	1	Max parallel feed fetches (max 16)
maxItems	number	—	Stop after N items are stored
analyzer	Function	—	Custom analyzer ({ item, context }) => AnalysisResult
model	AnalyzerConfig	—	Provider config for built-in LLM adapters

Returns FeedResultsArray — an Array<{ item, analysis }> with an extra .feedErrors property.

parser.parseURL(url, callback?)

Fetch a URL and return a raw parsed feed object (no analysis). Supports callback style.

parser.parseString(xml, callback?)

Parse raw XML string into a feed object. Synchronous internally.

parser.parseFile(path, callback?)

Read a local file and parse as feed XML.

---

runAgenticParser(config)

Low-level pipeline. Returns { results, feedErrors } directly.

runAgenticParser(config: AgenticParserConfig): Promise<AgenticParserResult>

AgenticParserConfig

Field	Type	Required	Description
feedUrls	string[]	✅	Feed URLs to process
dbPath	string	✅ (or storage)	SQLite DB path
storage	StorageAdapter	—	Custom adapter (overrides dbPath)
fetchFullArticle	boolean	—	Fetch full article body
concurrency	number	—	Max parallel feed fetches
maxItems	number	—	Stop after N items stored
parserOptions	ParserOptions	—	Options forwarded to XML parser
analyzer	Function	—	Custom analyzer function
model	AnalyzerConfig	—	Built-in provider config

Returns

{
  results: Array<{ item: NormalizedItem, analysis: AnalysisResult }>,
  feedErrors: Array<{ feedUrl: string, error: string }>
}

---

analyzeFeedItem(item, options?)

Analyse a single feed item. Used internally but exported for custom pipelines.

analyzeFeedItem(item, options?: {
  fetchFullArticle?: boolean;
  analyzer?: Function;
  signals?: string[];
  extraSignals?: string[];
  threshold?: number;
}): Promise<AnalysisResult>

---

heuristicAnalyze(item, context?, options?)

Synchronous heuristic scorer. No LLM, no async.

heuristicAnalyze(item, context?: string, options?: {
  signals?: string[];
  extraSignals?: string[];
  threshold?: number;   // default: 3
}): AnalysisResult

---

createStorage(dbPath)

Create a SQLite-backed StorageAdapter. Requires Node ≥ 22.5.0.

createStorage(dbPath: string): StorageAdapter

---

createMemoryStorage()

Create an in-memory StorageAdapter. Data is lost when the process exits. Works on any Node version.

createMemoryStorage(): StorageAdapter

---

createAnalyzer(config?)

Build an analyzer function for use in runAgenticParser({ analyzer }).

createAnalyzer(config?: AnalyzerConfig): Promise<AnalyzerFn>

AnalyzerConfig

Field	Type	Default	Description
provider	'heuristic'|'openai'|'anthropic'|'local'	'heuristic'	Analysis backend
model	string	provider default	Model ID string
apiKey	string	—	API key (required for openai/anthropic)
baseURL	string	—	Override endpoint URL
signals	string[]	—	Replace default heuristic signals
extraSignals	string[]	—	Append to default signals
threshold	number	3	Min signal matches to mark relevant

---

fetchFullArticle(url)

Fetch a URL and return its plain-text content (HTML stripped).

fetchFullArticle(url: string): Promise<string>

---

resolveSignals(options?)

Resolve the effective signal list given signals / extraSignals options.

resolveSignals(options?: { signals?: string[]; extraSignals?: string[] }): string[]

---

DEFAULT_HEURISTIC_SIGNALS

The built-in signal array used by the heuristic provider:

['release', 'security', 'vulnerability', 'node', 'javascript', 'typescript',
 'framework', 'api', 'breaking', 'performance', 'agent', 'rss']

---

AnalysisResult shape

{
  decision: 'relevant' | 'ignore';
  confidence: number;        // 0–100
  summary: string;
  impact: string;
  actionItems: string[];
  tags: string[];            // matched signals or LLM-assigned tags
}