capture

Documentation

Overview

Package capture fetches external URLs into note content behind SSRF guards. It is ported and trimmed from Seam v1 internal/capture: the URL path is kept (private-IP rejection, DNS-rebinding-safe pinned dialer, port allowlist, redirect-scheme and downgrade validation, response size cap); voice transcription is dropped.

Index

• Variables
• type URLContent
• type URLFetcher
  • func NewURLFetcher(allowedPorts []int) *URLFetcher
  • func (f *URLFetcher) FetchURL(ctx context.Context, rawURL string) (*URLContent, error)

Variables

var (
	ErrInvalidURL     = errors.New("invalid URL")
	ErrFetchFailed    = errors.New("URL fetch failed")
	ErrPrivateIP      = errors.New("URL points to private/loopback address")
	ErrUnsafeScheme   = errors.New("URL scheme not allowed")
	ErrDisallowedPort = errors.New("URL port not allowed")
)

Domain errors.

Types

type URLContent

type URLContent struct {
	Title string
	Body  string
	URL   string
}

URLContent holds the extracted content from a fetched URL.

type URLFetcher

type URLFetcher struct {
	// contains filtered or unexported fields
}

URLFetcher fetches and extracts readable content from URLs behind SSRF guards.

func NewURLFetcher

func NewURLFetcher(allowedPorts []int) *URLFetcher

NewURLFetcher creates a URLFetcher whose transport rejects private/loopback addresses and ports outside allowedPorts (connecting only to a validated IP, so DNS rebinding cannot slip past the check) and caps redirects at 10, each hop re-validated for an HTTP(S) scheme, no https->http downgrade, and -- because every hop dials through the same transport -- a public IP on an allowed port. An empty allowedPorts means the 80/443 default, never "any port".

func (*URLFetcher) FetchURL

func (f *URLFetcher) FetchURL(ctx context.Context, rawURL string) (*URLContent, error)

FetchURL fetches rawURL and extracts its title and main readable content. It rejects non-HTTP(S) schemes and empty hosts before dialing; the transport's SSRF guard rejects private addresses at connect time.