feat: first take at M4

This commit is contained in:
2026-04-10 15:32:05 +02:00
parent ee961f1b02
commit 3ca80b00d0
27 changed files with 3923 additions and 21 deletions

View File

@@ -0,0 +1,94 @@
use rusqlite::{params, Connection};
use crate::db::from_row::{generated_file_hash_from_row, GENERATED_FILE_HASH_COLUMNS};
use crate::model::GeneratedFileHash;
pub fn get_generated_file_hash(
conn: &Connection,
project_id: &str,
relative_path: &str,
) -> rusqlite::Result<GeneratedFileHash> {
conn.query_row(
&format!(
"SELECT {GENERATED_FILE_HASH_COLUMNS} FROM generated_file_hashes WHERE project_id = ?1 AND relative_path = ?2"
),
params![project_id, relative_path],
generated_file_hash_from_row,
)
}
pub fn upsert_generated_file_hash(conn: &Connection, hash: &GeneratedFileHash) -> rusqlite::Result<()> {
conn.execute(
"INSERT INTO generated_file_hashes (project_id, relative_path, content_hash, updated_at)
VALUES (?1, ?2, ?3, ?4)
ON CONFLICT(project_id, relative_path)
DO UPDATE SET content_hash = excluded.content_hash, updated_at = excluded.updated_at",
params![hash.project_id, hash.relative_path, hash.content_hash, hash.updated_at],
)?;
Ok(())
}
pub fn delete_generated_file_hash(
conn: &Connection,
project_id: &str,
relative_path: &str,
) -> rusqlite::Result<()> {
conn.execute(
"DELETE FROM generated_file_hashes WHERE project_id = ?1 AND relative_path = ?2",
params![project_id, relative_path],
)?;
Ok(())
}
pub fn list_generated_file_hashes_by_project(
conn: &Connection,
project_id: &str,
) -> rusqlite::Result<Vec<GeneratedFileHash>> {
let mut stmt = conn.prepare(&format!(
"SELECT {GENERATED_FILE_HASH_COLUMNS} FROM generated_file_hashes WHERE project_id = ?1 ORDER BY relative_path"
))?;
let rows = stmt.query_map(params![project_id], generated_file_hash_from_row)?;
rows.collect()
}
#[cfg(test)]
mod tests {
use super::*;
use crate::db::queries::project::{insert_project, make_test_project};
use crate::db::Database;
fn setup() -> Database {
let mut db = Database::open_in_memory().unwrap();
db.migrate().unwrap();
insert_project(db.conn(), &make_test_project("p1", "blog")).unwrap();
db
}
#[test]
fn upsert_and_get_generated_hash() {
let db = setup();
let hash = GeneratedFileHash {
project_id: "p1".into(),
relative_path: "index.html".into(),
content_hash: "abc".into(),
updated_at: 42,
};
upsert_generated_file_hash(db.conn(), &hash).unwrap();
let stored = get_generated_file_hash(db.conn(), "p1", "index.html").unwrap();
assert_eq!(stored.content_hash, "abc");
upsert_generated_file_hash(
db.conn(),
&GeneratedFileHash {
content_hash: "def".into(),
updated_at: 99,
..hash
},
)
.unwrap();
let stored = get_generated_file_hash(db.conn(), "p1", "index.html").unwrap();
assert_eq!(stored.content_hash, "def");
assert_eq!(stored.updated_at, 99);
}
}

View File

@@ -9,3 +9,4 @@ pub mod post_media;
pub mod template;
pub mod script;
pub mod setting;
pub mod generated_file_hash;

View File

@@ -0,0 +1,756 @@
use std::time::Duration;
use keyring::Entry;
use reqwest::blocking::Client;
use rusqlite::Connection;
use serde::{Deserialize, Serialize};
use serde_json::{json, Value};
use crate::db::queries::setting;
use crate::engine::{EngineError, EngineResult};
use crate::util::now_unix_ms;
const KEYRING_SERVICE: &str = "RuDS";
const KEYRING_SETTING_PREFIX: &str = "ai.endpoint";
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "lowercase")]
pub enum AiEndpointKind {
Online,
Airplane,
}
impl AiEndpointKind {
pub fn as_str(self) -> &'static str {
match self {
Self::Online => "online",
Self::Airplane => "airplane",
}
}
fn settings_prefix(self) -> String {
format!("ai.endpoint.{}", self.as_str())
}
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct AiEndpointConfig {
pub kind: AiEndpointKind,
pub url: String,
pub model: String,
pub api_key: Option<String>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct StoredAiEndpointConfig {
pub kind: AiEndpointKind,
pub url: String,
pub model: String,
pub api_key_configured: bool,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
pub struct AiSettings {
pub offline_mode: bool,
pub default_model: Option<String>,
pub title_model: Option<String>,
pub image_model: Option<String>,
pub system_prompt: String,
pub online_endpoint: StoredAiEndpointConfig,
pub airplane_endpoint: StoredAiEndpointConfig,
}
impl Default for StoredAiEndpointConfig {
fn default() -> Self {
Self {
kind: AiEndpointKind::Online,
url: String::new(),
model: String::new(),
api_key_configured: false,
}
}
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct AiModelInfo {
pub id: String,
pub name: String,
pub context_window: Option<u64>,
pub max_output_tokens: Option<u64>,
pub supports_vision: bool,
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum OneShotOperation {
AnalyzeTaxonomy,
AnalyzePost,
DetectLanguage,
TranslatePost { target_language: String },
AnalyzeImage,
TranslateMedia { target_language: String },
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct OneShotRequest {
pub operation: OneShotOperation,
pub content: Value,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct TaxonomySuggestion {
pub tags: Vec<String>,
pub categories: Vec<String>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct PostAnalysisResult {
pub title: String,
pub excerpt: String,
pub slug: String,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct LanguageDetectionResult {
pub language_code: String,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct TranslationResult {
pub title: String,
pub excerpt: String,
pub content: String,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct ImageAnalysisResult {
pub title: String,
pub alt: String,
pub caption: String,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct MediaTranslationResult {
pub title: String,
pub alt: String,
pub caption: String,
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum OneShotResponse {
Taxonomy(TaxonomySuggestion),
PostAnalysis(PostAnalysisResult),
LanguageDetection(LanguageDetectionResult),
Translation(TranslationResult),
ImageAnalysis(ImageAnalysisResult),
MediaTranslation(MediaTranslationResult),
}
pub fn load_ai_settings(conn: &Connection, offline_mode: bool) -> EngineResult<AiSettings> {
let online_endpoint = load_endpoint(conn, AiEndpointKind::Online)?;
let airplane_endpoint = load_endpoint(conn, AiEndpointKind::Airplane)?;
let default_model = get_optional_setting(conn, "ai.default_model")?;
let title_model = get_optional_setting(conn, "ai.title_model")?;
let image_model = get_optional_setting(conn, "ai.image_model")?;
let system_prompt = get_optional_setting(conn, "ai.system_prompt")?.unwrap_or_default();
Ok(AiSettings {
offline_mode,
default_model,
title_model,
image_model,
system_prompt,
online_endpoint,
airplane_endpoint,
})
}
pub fn save_endpoint(conn: &Connection, endpoint: &AiEndpointConfig) -> EngineResult<()> {
validate_endpoint_config(endpoint)?;
let checked_at = now_unix_ms();
set_setting(conn, &endpoint_setting_key(endpoint.kind, "url"), endpoint.url.trim(), checked_at)?;
set_setting(conn, &endpoint_setting_key(endpoint.kind, "model"), endpoint.model.trim(), checked_at)?;
if endpoint.kind == AiEndpointKind::Online {
let entry = endpoint_keyring_entry(endpoint.kind)?;
if let Some(api_key) = &endpoint.api_key {
if api_key.trim().is_empty() {
entry.delete_credential().ok();
set_setting(conn, &endpoint_setting_key(endpoint.kind, "api_key_configured"), "false", checked_at)?;
} else {
entry.set_password(api_key.trim()).map_err(keyring_error)?;
set_setting(conn, &endpoint_setting_key(endpoint.kind, "api_key_configured"), "true", checked_at)?;
}
}
}
Ok(())
}
pub fn save_model_preferences(
conn: &Connection,
default_model: Option<&str>,
title_model: Option<&str>,
image_model: Option<&str>,
system_prompt: &str,
) -> EngineResult<()> {
let updated_at = now_unix_ms();
set_optional_setting(conn, "ai.default_model", default_model, updated_at)?;
set_optional_setting(conn, "ai.title_model", title_model, updated_at)?;
set_optional_setting(conn, "ai.image_model", image_model, updated_at)?;
set_setting(conn, "ai.system_prompt", system_prompt, updated_at)?;
Ok(())
}
pub fn active_endpoint(conn: &Connection, offline_mode: bool) -> EngineResult<AiEndpointConfig> {
let kind = if offline_mode { AiEndpointKind::Airplane } else { AiEndpointKind::Online };
let stored = load_endpoint(conn, kind)?;
if stored.url.trim().is_empty() {
return Err(EngineError::Validation(format!(
"AI unavailable - configure {} endpoint in Settings",
kind.as_str()
)));
}
let api_key = if kind == AiEndpointKind::Online {
let entry = endpoint_keyring_entry(kind)?;
let password = entry.get_password().map_err(keyring_error)?;
if password.trim().is_empty() {
return Err(EngineError::Validation(
"AI unavailable - configure online endpoint in Settings".to_string(),
));
}
Some(password)
} else {
None
};
Ok(AiEndpointConfig {
kind,
url: stored.url,
model: stored.model,
api_key,
})
}
pub fn load_endpoint_api_key(kind: AiEndpointKind) -> EngineResult<Option<String>> {
let entry = endpoint_keyring_entry(kind)?;
match entry.get_password() {
Ok(password) if password.trim().is_empty() => Ok(None),
Ok(password) => Ok(Some(password)),
Err(keyring::Error::NoEntry) => Ok(None),
Err(error) => Err(keyring_error(error)),
}
}
pub fn refresh_model_catalog(endpoint: &AiEndpointConfig) -> EngineResult<Vec<AiModelInfo>> {
validate_endpoint_config(endpoint)?;
let client = build_http_client()?;
let request = client.get(models_url(&endpoint.url));
let response = with_auth(request, endpoint)
.send()?
.error_for_status()?;
let body: Value = response.json()?;
let models = body
.get("data")
.and_then(Value::as_array)
.ok_or_else(|| EngineError::Parse("model catalog response missing data array".to_string()))?;
let mut result = Vec::new();
for model in models {
let id = model
.get("id")
.and_then(Value::as_str)
.ok_or_else(|| EngineError::Parse("model entry missing id".to_string()))?
.to_string();
let name = model
.get("name")
.and_then(Value::as_str)
.unwrap_or(&id)
.to_string();
let context_window = model
.get("context_window")
.or_else(|| model.get("contextWindow"))
.and_then(Value::as_u64);
let max_output_tokens = model
.get("max_output_tokens")
.or_else(|| model.get("maxOutputTokens"))
.and_then(Value::as_u64);
let supports_vision = model
.get("modalities")
.and_then(Value::as_array)
.map(|modalities| modalities.iter().any(|value| value.as_str() == Some("vision")))
.unwrap_or(false);
result.push(AiModelInfo {
id,
name,
context_window,
max_output_tokens,
supports_vision,
});
}
Ok(result)
}
pub fn test_endpoint(endpoint: &AiEndpointConfig) -> EngineResult<()> {
let _ = refresh_model_catalog(endpoint)?;
Ok(())
}
pub fn run_one_shot(
conn: &Connection,
offline_mode: bool,
request: &OneShotRequest,
) -> EngineResult<OneShotResponse> {
let settings = load_ai_settings(conn, offline_mode)?;
let endpoint = active_endpoint(conn, offline_mode)?;
let model = select_model(&settings, &endpoint, &request.operation)?;
let prompt = build_one_shot_prompt(request)?;
let schema = response_schema(&request.operation);
let payload = json!({
"model": model,
"messages": [
{
"role": "system",
"content": build_system_prompt(&settings.system_prompt, &request.operation),
},
{
"role": "user",
"content": prompt,
}
],
"response_format": {
"type": "json_schema",
"json_schema": {
"name": schema.0,
"schema": schema.1,
"strict": true
}
}
});
let client = build_http_client()?;
let response = with_auth(client.post(chat_completions_url(&endpoint.url)).json(&payload), &endpoint)
.send()?
.error_for_status()?;
let body: Value = response.json()?;
let content = body
.get("choices")
.and_then(Value::as_array)
.and_then(|choices| choices.first())
.and_then(|choice| choice.get("message"))
.and_then(|message| message.get("content"))
.and_then(Value::as_str)
.ok_or_else(|| EngineError::Parse("chat completions response missing message content".to_string()))?;
parse_one_shot_response(request, content)
}
fn build_http_client() -> EngineResult<Client> {
Ok(Client::builder().timeout(Duration::from_secs(5)).build()?)
}
fn load_endpoint(conn: &Connection, kind: AiEndpointKind) -> EngineResult<StoredAiEndpointConfig> {
let url = get_optional_setting(conn, &endpoint_setting_key(kind, "url"))?.unwrap_or_default();
let model = get_optional_setting(conn, &endpoint_setting_key(kind, "model"))?.unwrap_or_default();
let api_key_configured = get_optional_setting(conn, &endpoint_setting_key(kind, "api_key_configured"))?
.map(|value| value == "true")
.unwrap_or(false);
Ok(StoredAiEndpointConfig {
kind,
url,
model,
api_key_configured,
})
}
fn validate_endpoint_config(endpoint: &AiEndpointConfig) -> EngineResult<()> {
if endpoint.url.trim().is_empty() {
return Err(EngineError::Validation("endpoint url is required".to_string()));
}
if endpoint.model.trim().is_empty() {
return Err(EngineError::Validation("endpoint model is required".to_string()));
}
if endpoint.kind == AiEndpointKind::Online
&& endpoint.api_key.as_ref().map(|value| value.trim().is_empty()).unwrap_or(true)
{
return Err(EngineError::Validation("online endpoint api key is required".to_string()));
}
Ok(())
}
fn select_model(settings: &AiSettings, endpoint: &AiEndpointConfig, operation: &OneShotOperation) -> EngineResult<String> {
let selected = match operation {
OneShotOperation::AnalyzeImage => settings.image_model.as_ref(),
OneShotOperation::AnalyzeTaxonomy
| OneShotOperation::AnalyzePost
| OneShotOperation::DetectLanguage
| OneShotOperation::TranslatePost { .. }
| OneShotOperation::TranslateMedia { .. } => settings.title_model.as_ref().or(settings.default_model.as_ref()),
}
.filter(|model| !model.trim().is_empty())
.cloned()
.unwrap_or_else(|| endpoint.model.clone());
if selected.trim().is_empty() {
return Err(EngineError::Validation("AI unavailable - configure model in Settings".to_string()));
}
Ok(selected)
}
fn build_system_prompt(base_prompt: &str, operation: &OneShotOperation) -> String {
let operation_prompt = match operation {
OneShotOperation::AnalyzeTaxonomy => "Return only JSON with tags and categories for the post.",
OneShotOperation::AnalyzePost => "Return only JSON with title, excerpt, and slug suggestions for the post.",
OneShotOperation::DetectLanguage => "Return only JSON with the detected language_code.",
OneShotOperation::TranslatePost { target_language } => {
return format!("{} Translate the post into {} and return only JSON with title, excerpt, and content.", base_prompt.trim(), target_language);
}
OneShotOperation::AnalyzeImage => "Return only JSON with title, alt, and caption suggestions for the image.",
OneShotOperation::TranslateMedia { target_language } => {
return format!("{} Translate the media metadata into {} and return only JSON with title, alt, and caption.", base_prompt.trim(), target_language);
}
};
if base_prompt.trim().is_empty() {
operation_prompt.to_string()
} else {
format!("{} {}", base_prompt.trim(), operation_prompt)
}
}
fn build_one_shot_prompt(request: &OneShotRequest) -> EngineResult<String> {
match &request.operation {
OneShotOperation::AnalyzeTaxonomy => Ok(format!(
"Suggest tags and categories for this post: {}",
serde_json::to_string(&request.content)?
)),
OneShotOperation::AnalyzePost => Ok(format!(
"Analyze this post and suggest title, excerpt, and slug: {}",
serde_json::to_string(&request.content)?
)),
OneShotOperation::DetectLanguage => Ok(format!(
"Detect the language of this text: {}",
serde_json::to_string(&request.content)?
)),
OneShotOperation::TranslatePost { target_language } => Ok(format!(
"Translate this post to {}: {}",
target_language,
serde_json::to_string(&request.content)?
)),
OneShotOperation::AnalyzeImage => Ok(format!(
"Analyze this image metadata and return title, alt, and caption suggestions: {}",
serde_json::to_string(&request.content)?
)),
OneShotOperation::TranslateMedia { target_language } => Ok(format!(
"Translate this media metadata to {}: {}",
target_language,
serde_json::to_string(&request.content)?
)),
}
}
fn response_schema(operation: &OneShotOperation) -> (&'static str, Value) {
match operation {
OneShotOperation::AnalyzeTaxonomy => (
"taxonomy_suggestion",
json!({
"type": "object",
"additionalProperties": false,
"properties": {
"tags": { "type": "array", "items": { "type": "string" } },
"categories": { "type": "array", "items": { "type": "string" } }
},
"required": ["tags", "categories"]
}),
),
OneShotOperation::AnalyzePost => (
"post_analysis",
json!({
"type": "object",
"additionalProperties": false,
"properties": {
"title": { "type": "string" },
"excerpt": { "type": "string" },
"slug": { "type": "string" }
},
"required": ["title", "excerpt", "slug"]
}),
),
OneShotOperation::DetectLanguage => (
"language_detection",
json!({
"type": "object",
"additionalProperties": false,
"properties": {
"language_code": { "type": "string" }
},
"required": ["language_code"]
}),
),
OneShotOperation::TranslatePost { .. } => (
"post_translation",
json!({
"type": "object",
"additionalProperties": false,
"properties": {
"title": { "type": "string" },
"excerpt": { "type": "string" },
"content": { "type": "string" }
},
"required": ["title", "excerpt", "content"]
}),
),
OneShotOperation::AnalyzeImage => (
"image_analysis",
json!({
"type": "object",
"additionalProperties": false,
"properties": {
"title": { "type": "string" },
"alt": { "type": "string" },
"caption": { "type": "string" }
},
"required": ["title", "alt", "caption"]
}),
),
OneShotOperation::TranslateMedia { .. } => (
"media_translation",
json!({
"type": "object",
"additionalProperties": false,
"properties": {
"title": { "type": "string" },
"alt": { "type": "string" },
"caption": { "type": "string" }
},
"required": ["title", "alt", "caption"]
}),
),
}
}
fn parse_one_shot_response(request: &OneShotRequest, content: &str) -> EngineResult<OneShotResponse> {
Ok(match request.operation {
OneShotOperation::AnalyzeTaxonomy => OneShotResponse::Taxonomy(serde_json::from_str(content)?),
OneShotOperation::AnalyzePost => OneShotResponse::PostAnalysis(serde_json::from_str(content)?),
OneShotOperation::DetectLanguage => OneShotResponse::LanguageDetection(serde_json::from_str(content)?),
OneShotOperation::TranslatePost { .. } => OneShotResponse::Translation(serde_json::from_str(content)?),
OneShotOperation::AnalyzeImage => OneShotResponse::ImageAnalysis(serde_json::from_str(content)?),
OneShotOperation::TranslateMedia { .. } => OneShotResponse::MediaTranslation(serde_json::from_str(content)?),
})
}
fn endpoint_setting_key(kind: AiEndpointKind, suffix: &str) -> String {
format!("{}.{}", kind.settings_prefix(), suffix)
}
fn endpoint_keyring_entry(kind: AiEndpointKind) -> EngineResult<Entry> {
Entry::new(KEYRING_SERVICE, &format!("{}.{}", KEYRING_SETTING_PREFIX, kind.as_str())).map_err(keyring_error)
}
fn keyring_error(error: keyring::Error) -> EngineError {
EngineError::Validation(error.to_string())
}
fn set_setting(conn: &Connection, key: &str, value: &str, updated_at: i64) -> EngineResult<()> {
setting::set_setting_value(conn, key, value, updated_at)?;
Ok(())
}
fn set_optional_setting(conn: &Connection, key: &str, value: Option<&str>, updated_at: i64) -> EngineResult<()> {
set_setting(conn, key, value.unwrap_or(""), updated_at)
}
fn get_optional_setting(conn: &Connection, key: &str) -> EngineResult<Option<String>> {
match setting::get_setting_by_key(conn, key) {
Ok(setting) if setting.value.trim().is_empty() => Ok(None),
Ok(setting) => Ok(Some(setting.value)),
Err(rusqlite::Error::QueryReturnedNoRows) => Ok(None),
Err(error) => Err(EngineError::Db(error)),
}
}
fn models_url(base_url: &str) -> String {
join_openai_path(base_url, "models")
}
fn chat_completions_url(base_url: &str) -> String {
join_openai_path(base_url, "chat/completions")
}
fn join_openai_path(base_url: &str, suffix: &str) -> String {
let trimmed = base_url.trim_end_matches('/');
if trimmed.ends_with("/v1") {
format!("{trimmed}/{suffix}")
} else {
format!("{trimmed}/v1/{suffix}")
}
}
fn with_auth(
request: reqwest::blocking::RequestBuilder,
endpoint: &AiEndpointConfig,
) -> reqwest::blocking::RequestBuilder {
if let Some(api_key) = &endpoint.api_key {
request.bearer_auth(api_key)
} else {
request
}
}
#[cfg(test)]
mod tests {
use std::io::{Read, Write};
use std::net::TcpListener;
use std::thread;
use super::*;
use crate::db::Database;
fn setup() -> Database {
let mut db = Database::open_in_memory().unwrap();
db.migrate().unwrap();
db
}
fn clear_keyring(kind: AiEndpointKind) {
let entry = endpoint_keyring_entry(kind).unwrap();
entry.delete_credential().ok();
}
#[test]
fn loads_empty_defaults() {
clear_keyring(AiEndpointKind::Online);
let db = setup();
let settings = load_ai_settings(db.conn(), false).unwrap();
assert!(!settings.offline_mode);
assert!(settings.online_endpoint.url.is_empty());
assert!(settings.airplane_endpoint.url.is_empty());
assert!(settings.default_model.is_none());
}
#[test]
fn saves_online_endpoint_with_keychain_secret() {
clear_keyring(AiEndpointKind::Online);
let db = setup();
save_endpoint(
db.conn(),
&AiEndpointConfig {
kind: AiEndpointKind::Online,
url: "https://example.test/v1".to_string(),
model: "gpt-4.1-mini".to_string(),
api_key: Some("secret-token".to_string()),
},
)
.unwrap();
let active = active_endpoint(db.conn(), false).unwrap();
assert_eq!(active.url, "https://example.test/v1");
assert_eq!(active.model, "gpt-4.1-mini");
assert_eq!(active.api_key.as_deref(), Some("secret-token"));
}
#[test]
fn airplane_endpoint_does_not_require_api_key() {
let db = setup();
save_endpoint(
db.conn(),
&AiEndpointConfig {
kind: AiEndpointKind::Airplane,
url: "http://localhost:11434/v1".to_string(),
model: "llama3.2".to_string(),
api_key: None,
},
)
.unwrap();
let active = active_endpoint(db.conn(), true).unwrap();
assert_eq!(active.kind, AiEndpointKind::Airplane);
assert!(active.api_key.is_none());
}
#[test]
fn refresh_model_catalog_parses_openai_models_shape() {
let server = spawn_test_server(|request| {
assert!(request.starts_with("GET /v1/models HTTP/1.1"));
http_ok(
r#"{"data":[{"id":"gpt-4.1-mini","name":"GPT 4.1 mini","context_window":128000,"max_output_tokens":8192,"modalities":["text"]},{"id":"gpt-4.1","modalities":["text","vision"]}]}"#,
)
});
let models = refresh_model_catalog(&AiEndpointConfig {
kind: AiEndpointKind::Airplane,
url: server,
model: "gpt-4.1-mini".to_string(),
api_key: None,
})
.unwrap();
assert_eq!(models.len(), 2);
assert_eq!(models[0].name, "GPT 4.1 mini");
assert!(models[1].supports_vision);
}
#[test]
fn run_one_shot_uses_active_endpoint_and_parses_response() {
clear_keyring(AiEndpointKind::Online);
let server = spawn_test_server(|request| {
if request.starts_with("GET /v1/models HTTP/1.1") {
return http_ok(r#"{"data":[{"id":"gpt-4.1-mini"}]}"#);
}
assert!(request.starts_with("POST /v1/chat/completions HTTP/1.1"));
assert!(request.contains("authorization: Bearer secret-token") || request.contains("Authorization: Bearer secret-token"));
http_ok(
r#"{"choices":[{"message":{"content":"{\"title\":\"Better title\",\"excerpt\":\"Short summary\",\"slug\":\"better-title\"}"}}]}"#,
)
});
let db = setup();
save_endpoint(
db.conn(),
&AiEndpointConfig {
kind: AiEndpointKind::Online,
url: server,
model: "gpt-4.1-mini".to_string(),
api_key: Some("secret-token".to_string()),
},
)
.unwrap();
save_model_preferences(db.conn(), None, Some("gpt-4.1-mini"), None, "").unwrap();
let response = run_one_shot(
db.conn(),
false,
&OneShotRequest {
operation: OneShotOperation::AnalyzePost,
content: json!({"title":"Draft title","excerpt":"","content":"Body"}),
},
)
.unwrap();
assert_eq!(
response,
OneShotResponse::PostAnalysis(PostAnalysisResult {
title: "Better title".to_string(),
excerpt: "Short summary".to_string(),
slug: "better-title".to_string(),
})
);
}
fn spawn_test_server(handler: impl Fn(String) -> String + Send + 'static) -> String {
let listener = TcpListener::bind("127.0.0.1:0").unwrap();
let addr = listener.local_addr().unwrap();
thread::spawn(move || {
for stream in listener.incoming().take(2) {
let mut stream = stream.unwrap();
let mut buffer = [0_u8; 8192];
let size = stream.read(&mut buffer).unwrap();
let request = String::from_utf8_lossy(&buffer[..size]).to_string();
let response = handler(request);
stream.write_all(response.as_bytes()).unwrap();
}
});
format!("http://{}", addr)
}
fn http_ok(body: &str) -> String {
format!(
"HTTP/1.1 200 OK\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
body.len(),
body
)
}
}

View File

@@ -46,6 +46,12 @@ impl From<std::io::Error> for EngineError {
}
}
impl From<reqwest::Error> for EngineError {
fn from(e: reqwest::Error) -> Self {
Self::Parse(e.to_string())
}
}
impl From<serde_json::Error> for EngineError {
fn from(e: serde_json::Error) -> Self {
Self::Parse(e.to_string())

View File

@@ -0,0 +1,227 @@
use std::path::Path;
use chrono::{DateTime, TimeZone, Utc};
use rusqlite::Connection;
use crate::engine::{EngineError, EngineResult};
use crate::model::{Post, ProjectMetadata};
use crate::render::{
GeneratedWriteOutcome, build_calendar_json, build_canonical_post_path,
render_markdown_to_html, render_starter_list_page, render_starter_single_post_page,
write_generated_file,
};
#[derive(Debug, Clone)]
pub struct PublishedPostSource {
pub post: Post,
pub body_markdown: String,
}
#[derive(Debug, Default, Clone)]
pub struct GenerationReport {
pub written_paths: Vec<String>,
pub skipped_paths: Vec<String>,
}
pub fn generate_starter_site(
conn: &Connection,
output_dir: &Path,
project_id: &str,
metadata: &ProjectMetadata,
posts: &[PublishedPostSource],
language: &str,
) -> EngineResult<GenerationReport> {
let mut report = GenerationReport::default();
let list_input = posts
.iter()
.map(|source| (source.post.clone(), source.body_markdown.clone()))
.collect::<Vec<_>>();
let index_page = render_starter_list_page(&list_input, metadata, language)
.map_err(|error| EngineError::Parse(error.to_string()))?;
write_out(conn, output_dir, project_id, &index_page.relative_path, &index_page.html, &mut report)?;
for source in posts {
let rendered = render_starter_single_post_page(&source.post, &source.body_markdown, metadata, language)
.map_err(|error| EngineError::Parse(error.to_string()))?;
write_out(conn, output_dir, project_id, &rendered.relative_path, &rendered.html, &mut report)?;
}
write_out(
conn,
output_dir,
project_id,
"calendar.json",
&build_calendar_json(&posts.iter().map(|source| source.post.clone()).collect::<Vec<_>>())?,
&mut report,
)?;
let rss = build_rss_xml(metadata, posts, language);
write_out(conn, output_dir, project_id, "rss.xml", &rss, &mut report)?;
write_out(conn, output_dir, project_id, "feed.xml", &rss, &mut report)?;
write_out(conn, output_dir, project_id, "atom.xml", &build_atom_xml(metadata, posts, language), &mut report)?;
write_out(conn, output_dir, project_id, "sitemap.xml", &build_sitemap_xml(metadata, posts, language), &mut report)?;
Ok(report)
}
fn write_out(
conn: &Connection,
output_dir: &Path,
project_id: &str,
relative_path: &str,
content: &str,
report: &mut GenerationReport,
) -> EngineResult<()> {
match write_generated_file(conn, output_dir, project_id, relative_path, content)
.map_err(|error| EngineError::Parse(error.to_string()))?
{
GeneratedWriteOutcome::Written => report.written_paths.push(relative_path.to_string()),
GeneratedWriteOutcome::SkippedUnchanged => report.skipped_paths.push(relative_path.to_string()),
}
Ok(())
}
fn build_rss_xml(metadata: &ProjectMetadata, posts: &[PublishedPostSource], language: &str) -> String {
let base_url = metadata.public_url.as_deref().unwrap_or("").trim_end_matches('/');
let last_build = posts
.iter()
.filter_map(|post| timestamp(post.post.published_at.unwrap_or(post.post.created_at)))
.max()
.unwrap_or_else(Utc::now);
let mut xml = vec![
"<?xml version=\"1.0\" encoding=\"UTF-8\"?>".to_string(),
"<rss version=\"2.0\" xmlns:content=\"http://purl.org/rss/1.0/modules/content/\" xmlns:dc=\"http://purl.org/dc/elements/1.1/\">".to_string(),
" <channel>".to_string(),
format!(" <title>{}</title>", escape_xml(&metadata.name)),
format!(" <link>{base_url}/</link>"),
format!(" <description>{}</description>", escape_xml(metadata.description.as_deref().unwrap_or(""))),
format!(" <lastBuildDate>{}</lastBuildDate>", last_build.format("%a, %d %b %Y %H:%M:%S GMT")),
" <generator>bDS</generator>".to_string(),
];
for source in posts {
let url = format!("{base_url}{}", build_canonical_post_path(&source.post, language, metadata.main_language.as_deref().unwrap_or("en")));
let published = timestamp(source.post.published_at.unwrap_or(source.post.created_at)).unwrap_or_else(Utc::now);
xml.push(" <item>".to_string());
xml.push(format!(" <title>{}</title>", escape_xml(&source.post.title)));
xml.push(format!(" <link>{url}</link>"));
xml.push(format!(" <guid isPermaLink=\"true\">{url}</guid>"));
xml.push(format!(" <pubDate>{}</pubDate>", published.format("%a, %d %b %Y %H:%M:%S GMT")));
if let Some(author) = &source.post.author {
xml.push(format!(" <author>{}</author>", escape_xml(author)));
}
xml.push(format!(" <dc:language>{}</dc:language>", escape_xml(language)));
let html = render_markdown_to_html(&source.body_markdown);
xml.push(format!(" <description><![CDATA[{html}]]></description>"));
xml.push(format!(" <content:encoded><![CDATA[{html}]]></content:encoded>"));
for category in &source.post.categories {
xml.push(format!(" <category>{}</category>", escape_xml(category)));
}
for tag in &source.post.tags {
xml.push(format!(" <category>{}</category>", escape_xml(tag)));
}
xml.push(" </item>".to_string());
}
xml.push(" </channel>".to_string());
xml.push("</rss>".to_string());
xml.join("\n")
}
fn build_atom_xml(metadata: &ProjectMetadata, posts: &[PublishedPostSource], language: &str) -> String {
let base_url = metadata.public_url.as_deref().unwrap_or("").trim_end_matches('/');
let updated = posts
.iter()
.filter_map(|post| timestamp(post.post.published_at.unwrap_or(post.post.created_at)))
.max()
.unwrap_or_else(Utc::now);
let mut xml = vec![
"<?xml version=\"1.0\" encoding=\"UTF-8\"?>".to_string(),
"<feed xmlns=\"http://www.w3.org/2005/Atom\">".to_string(),
format!(" <title>{}</title>", escape_xml(&metadata.name)),
format!(" <subtitle>{}</subtitle>", escape_xml(metadata.description.as_deref().unwrap_or(""))),
format!(" <id>{base_url}/</id>"),
format!(" <link href=\"{base_url}/\" rel=\"alternate\" />"),
format!(" <link href=\"{base_url}/atom.xml\" rel=\"self\" />"),
format!(" <updated>{}</updated>", updated.to_rfc3339_opts(chrono::SecondsFormat::Millis, true)),
];
for source in posts {
let url = format!("{base_url}{}", build_canonical_post_path(&source.post, language, metadata.main_language.as_deref().unwrap_or("en")));
let published = timestamp(source.post.published_at.unwrap_or(source.post.created_at)).unwrap_or_else(Utc::now);
let html = render_markdown_to_html(&source.body_markdown);
xml.push(format!(" <entry xml:lang=\"{}\">", escape_xml(language)));
xml.push(format!(" <title>{}</title>", escape_xml(&source.post.title)));
xml.push(format!(" <id>{url}</id>"));
xml.push(format!(" <link href=\"{url}\" />"));
xml.push(format!(" <updated>{}</updated>", published.to_rfc3339_opts(chrono::SecondsFormat::Millis, true)));
xml.push(format!(" <published>{}</published>", published.to_rfc3339_opts(chrono::SecondsFormat::Millis, true)));
if let Some(author) = &source.post.author {
xml.push(format!(" <author><name>{}</name></author>", escape_xml(author)));
}
xml.push(format!(" <summary type=\"xhtml\"><div xmlns=\"http://www.w3.org/1999/xhtml\">{html}</div></summary>"));
xml.push(format!(" <content type=\"xhtml\"><div xmlns=\"http://www.w3.org/1999/xhtml\">{html}</div></content>"));
for category in &source.post.categories {
xml.push(format!(" <category term=\"{}\" />", escape_xml(category)));
}
for tag in &source.post.tags {
xml.push(format!(" <category term=\"{}\" />", escape_xml(tag)));
}
xml.push(" </entry>".to_string());
}
xml.push("</feed>".to_string());
xml.join("\n")
}
fn build_sitemap_xml(metadata: &ProjectMetadata, posts: &[PublishedPostSource], language: &str) -> String {
let base_url = metadata.public_url.as_deref().unwrap_or("").trim_end_matches('/');
let index_lastmod = posts
.iter()
.filter_map(|post| timestamp(post.post.published_at.unwrap_or(post.post.created_at)))
.max()
.unwrap_or_else(Utc::now)
.to_rfc3339_opts(chrono::SecondsFormat::Millis, true);
let mut xml = vec![
"<?xml version=\"1.0\" encoding=\"UTF-8\"?>".to_string(),
"<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\" xmlns:xhtml=\"http://www.w3.org/1999/xhtml\">".to_string(),
" <url>".to_string(),
format!(" <loc>{base_url}/</loc>"),
format!(" <lastmod>{index_lastmod}</lastmod>"),
" <changefreq>daily</changefreq>".to_string(),
" <priority>1.0</priority>".to_string(),
" </url>".to_string(),
];
for source in posts {
let url = format!("{base_url}{}", build_canonical_post_path(&source.post, language, metadata.main_language.as_deref().unwrap_or("en")));
let lastmod = timestamp(source.post.published_at.unwrap_or(source.post.created_at))
.unwrap_or_else(Utc::now)
.to_rfc3339_opts(chrono::SecondsFormat::Millis, true);
xml.push(" <url>".to_string());
xml.push(format!(" <loc>{url}</loc>"));
xml.push(format!(" <lastmod>{lastmod}</lastmod>"));
xml.push(" <changefreq>weekly</changefreq>".to_string());
xml.push(" <priority>0.8</priority>".to_string());
xml.push(" </url>".to_string());
}
xml.push("</urlset>".to_string());
xml.join("\n")
}
fn timestamp(timestamp_ms: i64) -> Option<DateTime<Utc>> {
chrono::Utc.timestamp_millis_opt(timestamp_ms).single()
}
fn escape_xml(value: &str) -> String {
value
.replace('&', "&amp;")
.replace('<', "&lt;")
.replace('>', "&gt;")
.replace('"', "&quot;")
.replace('\'', "&apos;")
}

View File

@@ -1,4 +1,5 @@
pub mod error;
pub mod ai;
pub mod context;
pub mod project;
pub mod meta;
@@ -15,6 +16,8 @@ pub mod metadata_diff;
pub mod rebuild;
pub mod search;
pub mod calendar;
pub mod generation;
pub mod preview;
pub mod validate_translations;
pub mod validate_media;
pub mod validate_content;

View File

@@ -0,0 +1,126 @@
use crate::engine::{EngineError, EngineResult};
use crate::engine::generation::PublishedPostSource;
use crate::model::ProjectMetadata;
use crate::render::{build_canonical_post_path, render_starter_list_page, render_starter_single_post_page};
pub fn render_preview_path(
path: &str,
metadata: &ProjectMetadata,
posts: &[PublishedPostSource],
) -> EngineResult<Option<String>> {
let normalized = if path.is_empty() { "/" } else { path };
let main_language = metadata.main_language.as_deref().unwrap_or("en");
if normalized == "/" {
let list_posts = posts
.iter()
.map(|source| (source.post.clone(), source.body_markdown.clone()))
.collect::<Vec<_>>();
return render_starter_list_page(&list_posts, metadata, main_language)
.map(|page| Some(page.html))
.map_err(|error| EngineError::Parse(error.to_string()));
}
let (language, route_path) = split_language_prefix(normalized, metadata);
if let Some(source) = posts.iter().find(|source| {
build_canonical_post_path(&source.post, &language, main_language) == route_path
}) {
return render_starter_single_post_page(&source.post, &source.body_markdown, metadata, &language)
.map(|page| Some(page.html))
.map_err(|error| EngineError::Parse(error.to_string()));
}
Ok(None)
}
fn split_language_prefix(path: &str, metadata: &ProjectMetadata) -> (String, String) {
let trimmed = path.trim_start_matches('/');
let mut segments = trimmed.split('/');
let first = segments.next().unwrap_or_default();
if metadata.blog_languages.iter().any(|language| language == first) {
let remainder = segments.collect::<Vec<_>>().join("/");
return (first.to_string(), format!("/{first}/{}", remainder.trim_start_matches('/')));
}
(
metadata.main_language.as_deref().unwrap_or("en").to_string(),
path.to_string(),
)
}
#[cfg(test)]
mod tests {
use super::*;
use crate::model::{Post, PostStatus};
fn make_metadata() -> ProjectMetadata {
ProjectMetadata {
name: "Blog".into(),
description: None,
public_url: Some("https://example.com".into()),
main_language: Some("en".into()),
default_author: None,
max_posts_per_page: 50,
blogmark_category: None,
pico_theme: None,
semantic_similarity_enabled: false,
blog_languages: vec!["en".into(), "de".into()],
}
}
fn make_post() -> PublishedPostSource {
PublishedPostSource {
post: Post {
id: "post-1".into(),
project_id: "project-1".into(),
title: "Hello".into(),
slug: "hello".into(),
excerpt: None,
content: Some("Body".into()),
status: PostStatus::Published,
author: None,
language: Some("en".into()),
do_not_translate: false,
template_slug: None,
file_path: String::new(),
checksum: None,
tags: vec![],
categories: vec![],
published_title: None,
published_content: None,
published_tags: None,
published_categories: None,
published_excerpt: None,
created_at: 1_710_000_000_000,
updated_at: 1_710_000_000_000,
published_at: Some(1_710_000_000_000),
},
body_markdown: "Hello **world**".into(),
}
}
#[test]
fn root_preview_renders_index_page() {
let html = render_preview_path("/", &make_metadata(), &[make_post()])
.unwrap()
.unwrap();
assert!(html.contains("post-list"));
}
#[test]
fn preview_renders_single_post_for_canonical_path() {
let html = render_preview_path("/2024/03/09/hello", &make_metadata(), &[make_post()])
.unwrap()
.unwrap();
assert!(html.contains("<h1>Hello</h1>"));
assert!(html.contains("<strong>world</strong>"));
}
#[test]
fn preview_renders_language_prefixed_single_post() {
let html = render_preview_path("/de/2024/03/09/hello", &make_metadata(), &[make_post()])
.unwrap()
.unwrap();
assert!(html.contains("lang=\"de\""));
}
}

View File

@@ -0,0 +1,104 @@
use std::collections::BTreeMap;
use std::fs;
use std::path::Path;
use chrono::{Datelike, TimeZone, Utc};
use rusqlite::Connection;
use serde::Serialize;
use crate::db::queries::generated_file_hash as qhash;
use crate::model::{GeneratedFileHash, Post};
use crate::util::{atomic_write_str, content_hash, now_unix_ms};
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum GeneratedWriteOutcome {
Written,
SkippedUnchanged,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize)]
pub struct CalendarArchiveData {
pub years: BTreeMap<String, usize>,
pub months: BTreeMap<String, usize>,
pub days: BTreeMap<String, usize>,
}
pub fn write_generated_file(
conn: &Connection,
output_dir: &Path,
project_id: &str,
relative_path: &str,
content: &str,
) -> Result<GeneratedWriteOutcome, Box<dyn std::error::Error + Send + Sync>> {
let hash = content_hash(content.as_bytes());
if let Ok(existing) = qhash::get_generated_file_hash(conn, project_id, relative_path) {
if existing.content_hash == hash {
return Ok(GeneratedWriteOutcome::SkippedUnchanged);
}
}
let target_path = output_dir.join(relative_path);
if let Some(parent) = target_path.parent() {
fs::create_dir_all(parent)?;
}
atomic_write_str(&target_path, content)?;
qhash::upsert_generated_file_hash(
conn,
&GeneratedFileHash {
project_id: project_id.to_string(),
relative_path: relative_path.to_string(),
content_hash: hash,
updated_at: now_unix_ms(),
},
)?;
Ok(GeneratedWriteOutcome::Written)
}
pub fn build_core_generation_paths(main_language: &str, blog_languages: &[String]) -> Vec<String> {
let mut paths = vec![
"index.html".to_string(),
"sitemap.xml".to_string(),
"feed.xml".to_string(),
"atom.xml".to_string(),
"calendar.json".to_string(),
];
for language in blog_languages {
if language != main_language {
paths.push(format!("{language}/index.html"));
paths.push(format!("{language}/feed.xml"));
paths.push(format!("{language}/atom.xml"));
}
}
paths
}
pub fn build_calendar_archive_data(posts: &[Post]) -> CalendarArchiveData {
let mut years = BTreeMap::new();
let mut months = BTreeMap::new();
let mut days = BTreeMap::new();
for post in posts {
let timestamp_ms = post.published_at.unwrap_or(post.created_at);
let Some(created_at) = Utc.timestamp_millis_opt(timestamp_ms).single() else {
continue;
};
let year = created_at.year().to_string();
let month = format!("{year}-{:02}", created_at.month());
let day = format!("{month}-{:02}", created_at.day());
*years.entry(year).or_insert(0) += 1;
*months.entry(month).or_insert(0) += 1;
*days.entry(day).or_insert(0) += 1;
}
CalendarArchiveData { years, months, days }
}
pub fn build_calendar_json(posts: &[Post]) -> serde_json::Result<String> {
serde_json::to_string_pretty(&build_calendar_archive_data(posts))
}

View File

@@ -0,0 +1,25 @@
use pulldown_cmark::{Options, Parser, html};
pub fn render_markdown_to_html(markdown: &str) -> String {
let mut options = Options::empty();
options.insert(Options::ENABLE_STRIKETHROUGH);
options.insert(Options::ENABLE_TABLES);
options.insert(Options::ENABLE_TASKLISTS);
let parser = Parser::new_ext(markdown, options);
let mut output = String::new();
html::push_html(&mut output, parser);
output
}
#[cfg(test)]
mod tests {
use super::render_markdown_to_html;
#[test]
fn renders_commonmark_to_html() {
let rendered = render_markdown_to_html("# Title\n\nSome *text*.");
assert!(rendered.contains("<h1>Title</h1>"));
assert!(rendered.contains("<p>Some <em>text</em>.</p>"));
}
}

View File

@@ -1 +1,20 @@
// Rendering pipeline — stubs for M0, implemented in M4.
mod markdown;
mod generation;
mod page_renderer;
mod routes;
mod template_lookup;
pub use generation::{
CalendarArchiveData, GeneratedWriteOutcome, build_calendar_json,
build_core_generation_paths, write_generated_file,
};
pub use markdown::render_markdown_to_html;
pub use page_renderer::{RenderError, render_liquid_template};
pub use routes::{
RenderedPage, build_canonical_post_path, render_starter_list_page,
render_starter_single_post_page,
};
pub use template_lookup::{
RenderCategorySettings, RenderTemplateLookup, TemplateLookupError,
resolve_post_template,
};

View File

@@ -0,0 +1,335 @@
use std::collections::HashMap;
use liquid::ParserBuilder;
use liquid::partials::{EagerCompiler, InMemorySource};
use liquid_core::{
Display_filter, Expression, Filter, FilterParameters, FilterReflection,
FromFilterParameters, ParseFilter, Runtime, Value, ValueView,
};
use serde::Serialize;
use thiserror::Error;
use crate::i18n::translate_render;
use crate::render::render_markdown_to_html;
#[derive(Debug, Error)]
pub enum RenderError {
#[error("liquid error: {0}")]
Liquid(#[from] liquid::Error),
}
#[derive(Debug, Clone, Default)]
struct HtmlRewriteContext {
canonical_post_path_by_slug: HashMap<String, String>,
canonical_media_path_by_source_path: HashMap<String, String>,
}
pub fn render_liquid_template<T: Serialize>(
template_source: &str,
partials: &HashMap<String, String>,
context: &T,
) -> Result<String, RenderError> {
let mut compiled_partials: EagerCompiler<InMemorySource> = EagerCompiler::empty();
for (name, content) in partials {
compiled_partials.add(format!("{name}.liquid"), content.clone());
}
let parser = ParserBuilder::with_stdlib()
.filter(I18n)
.filter(Markdown)
.partials(compiled_partials)
.build()?;
let template = parser.parse(template_source)?;
let globals = liquid::to_object(context)?;
Ok(template.render(&globals)?)
}
#[derive(Debug, FilterParameters)]
struct I18nArgs {
#[parameter(description = "Render language", arg_type = "str")]
language: Expression,
}
#[derive(Clone, ParseFilter, FilterReflection)]
#[filter(
name = "i18n",
description = "Translate a render key for a content language.",
parameters(I18nArgs),
parsed(I18nFilter)
)]
struct I18n;
#[derive(Debug, FromFilterParameters, Display_filter)]
#[name = "i18n"]
struct I18nFilter {
#[parameters]
args: I18nArgs,
}
impl Filter for I18nFilter {
fn evaluate(&self, input: &dyn ValueView, runtime: &dyn Runtime) -> liquid_core::Result<Value> {
let args = self.args.evaluate(runtime)?;
let key = input.to_kstr();
let language = args.language.to_kstr();
Ok(Value::scalar(translate_render(language.as_str(), key.as_str())))
}
}
#[derive(Debug, FilterParameters)]
struct MarkdownArgs {
#[parameter(description = "Post id", arg_type = "str")]
post_id: Option<Expression>,
#[parameter(description = "Post data by id", arg_type = "any")]
post_data_json_by_id: Option<Expression>,
#[parameter(description = "Canonical post path map", arg_type = "any")]
canonical_post_path_by_slug: Option<Expression>,
#[parameter(description = "Canonical media path map", arg_type = "any")]
canonical_media_path_by_source_path: Option<Expression>,
#[parameter(description = "Render language", arg_type = "str")]
language: Option<Expression>,
#[parameter(description = "Language prefix", arg_type = "str")]
language_prefix: Option<Expression>,
}
#[derive(Clone, ParseFilter, FilterReflection)]
#[filter(
name = "markdown",
description = "Render markdown to HTML and rewrite preview URLs.",
parameters(MarkdownArgs),
parsed(MarkdownFilter)
)]
struct Markdown;
#[derive(Debug, FromFilterParameters, Display_filter)]
#[name = "markdown"]
struct MarkdownFilter {
#[parameters]
args: MarkdownArgs,
}
impl Filter for MarkdownFilter {
fn evaluate(&self, input: &dyn ValueView, runtime: &dyn Runtime) -> liquid_core::Result<Value> {
let args = self.args.evaluate(runtime)?;
let markdown = input.to_kstr();
let rewrite_context = HtmlRewriteContext {
canonical_post_path_by_slug: args
.canonical_post_path_by_slug
.as_ref()
.map(value_to_string_map)
.unwrap_or_default(),
canonical_media_path_by_source_path: args
.canonical_media_path_by_source_path
.as_ref()
.map(value_to_string_map)
.unwrap_or_default(),
};
let rendered = render_markdown_to_html(markdown.as_str());
Ok(Value::scalar(rewrite_rendered_html_urls(&rendered, &rewrite_context)))
}
}
fn value_to_string_map(value: &impl ValueView) -> HashMap<String, String> {
value
.as_object()
.map(|object| {
object
.iter()
.map(|(key, value)| (key.to_string(), value.to_kstr().into_owned().to_string()))
.collect()
})
.unwrap_or_default()
}
pub(crate) fn rewrite_rendered_html_urls(html: &str, rewrite_context: &impl RewriteContextView) -> String {
let rewritten = rewrite_attribute_urls(html, "href", |href| normalize_preview_href(href, rewrite_context));
rewrite_attribute_urls(&rewritten, "src", |src| normalize_preview_src(src, rewrite_context))
}
pub(crate) trait RewriteContextView {
fn canonical_post_path_by_slug(&self) -> &HashMap<String, String>;
fn canonical_media_path_by_source_path(&self) -> &HashMap<String, String>;
}
impl RewriteContextView for HtmlRewriteContext {
fn canonical_post_path_by_slug(&self) -> &HashMap<String, String> {
&self.canonical_post_path_by_slug
}
fn canonical_media_path_by_source_path(&self) -> &HashMap<String, String> {
&self.canonical_media_path_by_source_path
}
}
fn rewrite_attribute_urls(
html: &str,
attribute: &str,
normalize: impl Fn(&str) -> String,
) -> String {
let mut result = String::with_capacity(html.len());
let mut cursor = 0;
while let Some(offset) = html[cursor..].find(attribute) {
let attr_start = cursor + offset;
result.push_str(&html[cursor..attr_start]);
result.push_str(attribute);
let after_attr = attr_start + attribute.len();
if !html[after_attr..].starts_with('=') {
cursor = after_attr;
continue;
}
result.push('=');
let quote_index = after_attr + 1;
let Some(quote) = html[quote_index..].chars().next() else {
cursor = after_attr + 1;
continue;
};
if quote != '\'' && quote != '"' {
cursor = after_attr + 1;
continue;
}
result.push(quote);
let value_start = quote_index + quote.len_utf8();
let Some(value_end_rel) = html[value_start..].find(quote) else {
result.push_str(&html[value_start..]);
return result;
};
let value_end = value_start + value_end_rel;
result.push_str(&normalize(&html[value_start..value_end]));
result.push(quote);
cursor = value_end + quote.len_utf8();
}
result.push_str(&html[cursor..]);
result
}
fn normalize_preview_href(raw_href: &str, rewrite_context: &impl RewriteContextView) -> String {
if raw_href.is_empty() || is_external_or_special_url(raw_href) {
return raw_href.to_string();
}
let (path_part, suffix) = split_path_suffix(raw_href.trim());
if let Some(normalized) = normalize_day_route(path_part) {
return format!("{normalized}{suffix}");
}
if let Some(slug) = extract_post_slug(path_part) {
let canonical = rewrite_context
.canonical_post_path_by_slug()
.get(&slug)
.cloned()
.unwrap_or_else(|| format!("/posts/{slug}"));
return format!("{canonical}{suffix}");
}
if let Some(media_source_key) = extract_media_source_key(path_part) {
let canonical = rewrite_context
.canonical_media_path_by_source_path()
.get(&media_source_key)
.cloned()
.unwrap_or_else(|| format!("/{media_source_key}"));
return format!("{canonical}{suffix}");
}
raw_href.to_string()
}
fn normalize_preview_src(raw_src: &str, rewrite_context: &impl RewriteContextView) -> String {
if raw_src.is_empty() || is_external_or_special_url(raw_src) {
return raw_src.to_string();
}
let (path_part, suffix) = split_path_suffix(raw_src.trim());
if let Some(media_source_key) = extract_media_source_key(path_part) {
let canonical = rewrite_context
.canonical_media_path_by_source_path()
.get(&media_source_key)
.cloned()
.unwrap_or_else(|| format!("/{media_source_key}"));
return format!("{canonical}{suffix}");
}
raw_src.to_string()
}
fn is_external_or_special_url(value: &str) -> bool {
let normalized = value.trim();
if normalized.is_empty() {
return false;
}
if normalized.starts_with('#') || normalized.starts_with("//") {
return true;
}
let mut seen_alpha = false;
for ch in normalized.chars() {
if ch == ':' {
return seen_alpha;
}
if ch.is_ascii_alphanumeric() || matches!(ch, '+' | '.' | '-') {
seen_alpha = true;
continue;
}
break;
}
false
}
fn split_path_suffix(value: &str) -> (&str, &str) {
let split_index = value.find(['?', '#']).unwrap_or(value.len());
(&value[..split_index], &value[split_index..])
}
fn normalize_day_route(path: &str) -> Option<String> {
let segments: Vec<_> = path.trim_start_matches('/').split('/').collect();
if segments.len() != 4 {
return None;
}
let [year, month, day, slug] = segments.as_slice() else {
return None;
};
if year.len() != 4 || !year.chars().all(|ch| ch.is_ascii_digit()) {
return None;
}
let month = month.parse::<u32>().ok()?;
let day = day.parse::<u32>().ok()?;
if slug.is_empty() {
return None;
}
Some(format!("/{year}/{month:02}/{day:02}/{slug}"))
}
fn extract_post_slug(path: &str) -> Option<String> {
let trimmed = path.trim_start_matches('/');
let segments: Vec<_> = trimmed.split('/').collect();
match segments.as_slice() {
["post" | "posts", slug] => Some(trim_html_suffix(slug)),
["post" | "posts", year, month, slug] if year.len() == 4 && month.chars().all(|ch| ch.is_ascii_digit()) => {
Some(trim_html_suffix(slug))
}
_ => None,
}
}
fn extract_media_source_key(path: &str) -> Option<String> {
let trimmed = path.trim_start_matches('/');
let segments: Vec<_> = trimmed.split('/').collect();
match segments.as_slice() {
["media", year, month, filename] if year.len() == 4 && month.len() == 2 => {
Some(format!("media/{year}/{month}/{}", filename.to_lowercase()))
}
_ => None,
}
}
fn trim_html_suffix(value: &str) -> String {
value
.trim_end_matches(".html")
.trim_end_matches(".htm")
.to_string()
}

View File

@@ -0,0 +1,474 @@
use std::collections::HashMap;
use chrono::{Datelike, TimeZone, Utc};
use serde::Serialize;
use crate::i18n::normalize_language;
use crate::model::{Post, ProjectMetadata};
use crate::render::{RenderError, render_liquid_template};
const STARTER_SINGLE_POST_TEMPLATE: &str = include_str!("../../../../assets/starter-templates/single-post.liquid");
const STARTER_POST_LIST_TEMPLATE: &str = include_str!("../../../../assets/starter-templates/post-list.liquid");
const STARTER_HEAD_PARTIAL: &str = include_str!("../../../../assets/starter-templates/partials/head.liquid");
const STARTER_MENU_PARTIAL: &str = include_str!("../../../../assets/starter-templates/partials/menu.liquid");
const STARTER_LANGUAGE_SWITCHER_PARTIAL: &str = include_str!("../../../../assets/starter-templates/partials/language-switcher.liquid");
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct RenderedPage {
pub relative_path: String,
pub html: String,
}
#[derive(Debug, Clone, Serialize)]
struct AlternateLinkContext {
href: String,
hreflang: String,
}
#[derive(Debug, Clone, Serialize)]
struct BlogLanguageContext {
is_current: bool,
code: String,
flag: String,
href: String,
href_prefix: String,
}
#[derive(Debug, Clone, Serialize)]
struct DayBlockContext {
show_date_marker: bool,
date_label: String,
posts: Vec<serde_json::Value>,
show_separator: bool,
}
#[derive(Debug, Clone, Serialize)]
struct ListTemplateContext {
language: String,
language_prefix: String,
page_title: String,
pico_stylesheet_href: Option<String>,
html_theme_attribute: Option<String>,
alternate_links: Vec<AlternateLinkContext>,
blog_languages: Vec<BlogLanguageContext>,
menu_items: Vec<serde_json::Value>,
calendar_initial_year: i32,
calendar_initial_month: u32,
archive_context: Option<serde_json::Value>,
show_archive_range_heading: bool,
min_date: Option<serde_json::Value>,
max_date: Option<serde_json::Value>,
day_blocks: Vec<DayBlockContext>,
is_list_page: bool,
is_first_page: bool,
is_last_page: bool,
has_prev_page: bool,
has_next_page: bool,
prev_page_href: Option<String>,
next_page_href: Option<String>,
canonical_post_path_by_slug: HashMap<String, String>,
canonical_media_path_by_source_path: HashMap<String, String>,
post_data_json_by_id: HashMap<String, String>,
}
#[derive(Debug, Clone, Serialize)]
struct PostTemplateContext<'a> {
language: &'a str,
language_prefix: String,
page_title: &'a str,
pico_stylesheet_href: Option<String>,
html_theme_attribute: Option<String>,
alternate_links: Vec<AlternateLinkContext>,
blog_languages: Vec<BlogLanguageContext>,
menu_items: Vec<serde_json::Value>,
calendar_initial_year: i32,
calendar_initial_month: u32,
post: serde_json::Value,
post_categories: Vec<String>,
post_tags: Vec<String>,
tag_color_by_name: HashMap<String, String>,
backlinks: Vec<serde_json::Value>,
canonical_post_path_by_slug: HashMap<String, String>,
canonical_media_path_by_source_path: HashMap<String, String>,
post_data_json_by_id: HashMap<String, String>,
}
pub fn build_canonical_post_path(post: &Post, language: &str, main_language: &str) -> String {
let timestamp_ms = post.published_at.unwrap_or(post.created_at);
let Some(timestamp) = Utc.timestamp_millis_opt(timestamp_ms).single() else {
return fallback_language_path(post, language, main_language);
};
let base = format!(
"/{:04}/{:02}/{:02}/{}",
timestamp.year(),
timestamp.month(),
timestamp.day(),
post.slug
);
if language.eq_ignore_ascii_case(main_language) {
base
} else {
format!("/{language}{base}")
}
}
pub fn render_starter_single_post_page(
post: &Post,
body_markdown: &str,
metadata: &ProjectMetadata,
language: &str,
) -> Result<RenderedPage, RenderError> {
let relative_path = format!("{}/index.html", build_canonical_post_path(post, language, main_language(metadata)).trim_start_matches('/'));
let canonical_path = build_canonical_post_path(post, language, main_language(metadata));
let (calendar_initial_year, calendar_initial_month) = calendar_initial_parts(post);
let context = PostTemplateContext {
language,
language_prefix: language_prefix(language, main_language(metadata)),
page_title: &post.title,
pico_stylesheet_href: metadata
.pico_theme
.as_ref()
.map(|_| "/assets/pico.min.css".to_string()),
html_theme_attribute: None,
alternate_links: build_alternate_links(post, metadata, language),
blog_languages: build_blog_languages(post, metadata, language),
menu_items: vec![],
calendar_initial_year,
calendar_initial_month,
post: serde_json::json!({
"id": post.id,
"title": post.title,
"content": body_markdown,
}),
post_categories: post.categories.clone(),
post_tags: post.tags.clone(),
tag_color_by_name: post
.tags
.iter()
.map(|tag| (tag.clone(), String::new()))
.collect(),
backlinks: vec![],
canonical_post_path_by_slug: HashMap::from([(post.slug.clone(), canonical_path)]),
canonical_media_path_by_source_path: HashMap::new(),
post_data_json_by_id: HashMap::new(),
};
let html = render_liquid_template(
STARTER_SINGLE_POST_TEMPLATE,
&starter_partials(),
&context,
)?;
Ok(RenderedPage { relative_path, html })
}
pub fn render_starter_list_page(
posts: &[(Post, String)],
metadata: &ProjectMetadata,
language: &str,
) -> Result<RenderedPage, RenderError> {
let relative_path = if language.eq_ignore_ascii_case(main_language(metadata)) {
"index.html".to_string()
} else {
format!("{language}/index.html")
};
let canonical_paths = posts
.iter()
.map(|(post, _)| {
(
post.slug.clone(),
build_canonical_post_path(post, language, main_language(metadata)),
)
})
.collect::<HashMap<_, _>>();
let (calendar_initial_year, calendar_initial_month) = posts
.first()
.map(|(post, _)| calendar_initial_parts(post))
.unwrap_or((1970, 1));
let context = ListTemplateContext {
language: language.to_string(),
language_prefix: language_prefix(language, main_language(metadata)),
page_title: metadata.name.clone(),
pico_stylesheet_href: metadata
.pico_theme
.as_ref()
.map(|_| "/assets/pico.min.css".to_string()),
html_theme_attribute: None,
alternate_links: vec![],
blog_languages: build_blog_languages_for_index(metadata, language),
menu_items: vec![],
calendar_initial_year,
calendar_initial_month,
archive_context: None,
show_archive_range_heading: false,
min_date: None,
max_date: None,
day_blocks: build_day_blocks(posts),
is_list_page: false,
is_first_page: true,
is_last_page: true,
has_prev_page: false,
has_next_page: false,
prev_page_href: None,
next_page_href: None,
canonical_post_path_by_slug: canonical_paths,
canonical_media_path_by_source_path: HashMap::new(),
post_data_json_by_id: HashMap::new(),
};
let html = render_liquid_template(
&starter_post_list_template(),
&starter_partials(),
&context,
)?;
Ok(RenderedPage { relative_path, html })
}
fn starter_partials() -> HashMap<String, String> {
HashMap::from([
("partials/head".to_string(), STARTER_HEAD_PARTIAL.to_string()),
("partials/menu".to_string(), STARTER_MENU_PARTIAL.to_string()),
(
"partials/language-switcher".to_string(),
STARTER_LANGUAGE_SWITCHER_PARTIAL.to_string(),
),
(
"partials/menu-items".to_string(),
"{% for item in items %}<a href=\"{{ item.href }}\">{{ item.title }}</a>{% endfor %}".to_string(),
),
])
}
fn starter_post_list_template() -> String {
STARTER_POST_LIST_TEMPLATE.replace(
"{% render 'partials/head', page_title: page_title, pico_stylesheet_href: pico_stylesheet_href, language_prefix: language_prefix %}",
"{% render 'partials/head', page_title: page_title, pico_stylesheet_href: pico_stylesheet_href, language_prefix: language_prefix, alternate_links: alternate_links %}",
)
}
fn main_language(metadata: &ProjectMetadata) -> &str {
metadata.main_language.as_deref().unwrap_or("en")
}
fn language_prefix(language: &str, main_language: &str) -> String {
if language.eq_ignore_ascii_case(main_language) {
String::new()
} else {
format!("/{language}")
}
}
fn fallback_language_path(post: &Post, language: &str, main_language: &str) -> String {
if language.eq_ignore_ascii_case(main_language) {
format!("/posts/{}", post.slug)
} else {
format!("/{language}/posts/{}", post.slug)
}
}
fn calendar_initial_parts(post: &Post) -> (i32, u32) {
let timestamp_ms = post.published_at.unwrap_or(post.created_at);
Utc.timestamp_millis_opt(timestamp_ms)
.single()
.map(|timestamp| (timestamp.year(), timestamp.month()))
.unwrap_or((1970, 1))
}
fn build_alternate_links(post: &Post, metadata: &ProjectMetadata, current_language: &str) -> Vec<AlternateLinkContext> {
metadata
.blog_languages
.iter()
.map(|language| AlternateLinkContext {
href: build_absolute_post_url(post, metadata, language),
hreflang: language.clone(),
})
.chain(std::iter::once(AlternateLinkContext {
href: build_absolute_post_url(post, metadata, current_language),
hreflang: "x-default".to_string(),
}))
.collect()
}
fn build_blog_languages(post: &Post, metadata: &ProjectMetadata, current_language: &str) -> Vec<BlogLanguageContext> {
metadata
.blog_languages
.iter()
.map(|language| BlogLanguageContext {
is_current: language.eq_ignore_ascii_case(current_language),
code: language.clone(),
flag: render_flag(language),
href: build_absolute_post_url(post, metadata, language),
href_prefix: language_prefix(language, main_language(metadata)),
})
.collect()
}
fn build_blog_languages_for_index(metadata: &ProjectMetadata, current_language: &str) -> Vec<BlogLanguageContext> {
metadata
.blog_languages
.iter()
.map(|language| BlogLanguageContext {
is_current: language.eq_ignore_ascii_case(current_language),
code: language.clone(),
flag: render_flag(language),
href: build_absolute_index_url(metadata, language),
href_prefix: language_prefix(language, main_language(metadata)),
})
.collect()
}
fn build_absolute_post_url(post: &Post, metadata: &ProjectMetadata, language: &str) -> String {
let base_url = metadata.public_url.as_deref().unwrap_or("").trim_end_matches('/');
format!("{base_url}{}", build_canonical_post_path(post, language, main_language(metadata)))
}
fn build_absolute_index_url(metadata: &ProjectMetadata, language: &str) -> String {
let base_url = metadata.public_url.as_deref().unwrap_or("").trim_end_matches('/');
let suffix = if language.eq_ignore_ascii_case(main_language(metadata)) {
"/".to_string()
} else {
format!("/{language}/")
};
format!("{base_url}{suffix}")
}
fn render_flag(language: &str) -> String {
normalize_language(language).flag_emoji().to_string()
}
fn build_day_blocks(posts: &[(Post, String)]) -> Vec<DayBlockContext> {
let mut blocks: Vec<DayBlockContext> = Vec::new();
let mut current_key: Option<String> = None;
for (post, body) in posts {
let timestamp_ms = post.published_at.unwrap_or(post.created_at);
let Some(timestamp) = Utc.timestamp_millis_opt(timestamp_ms).single() else {
continue;
};
let key = format!("{:04}-{:02}-{:02}", timestamp.year(), timestamp.month(), timestamp.day());
if current_key.as_deref() != Some(key.as_str()) {
if let Some(last) = blocks.last_mut() {
last.show_separator = true;
}
current_key = Some(key);
blocks.push(DayBlockContext {
show_date_marker: true,
date_label: format!("{:02}.{:02}.{:04}", timestamp.day(), timestamp.month(), timestamp.year()),
posts: Vec::new(),
show_separator: false,
});
}
if let Some(block) = blocks.last_mut() {
block.posts.push(serde_json::json!({
"id": post.id,
"slug": post.slug,
"title": post.title,
"content": post.excerpt.clone().unwrap_or_else(|| body.clone()),
"show_title": true,
}));
}
}
blocks
}
#[cfg(test)]
mod tests {
use super::*;
use crate::model::PostStatus;
fn make_post() -> Post {
Post {
id: "post-1".into(),
project_id: "project-1".into(),
title: "Hello".into(),
slug: "hello".into(),
excerpt: None,
content: Some("Body".into()),
status: PostStatus::Published,
author: None,
language: Some("en".into()),
do_not_translate: false,
template_slug: None,
file_path: String::new(),
checksum: None,
tags: vec![],
categories: vec![],
published_title: None,
published_content: None,
published_tags: None,
published_categories: None,
published_excerpt: None,
created_at: 1_710_000_000_000,
updated_at: 1_710_000_000_000,
published_at: Some(1_710_000_000_000),
}
}
fn make_metadata() -> ProjectMetadata {
ProjectMetadata {
name: "Blog".into(),
description: None,
public_url: Some("https://example.com".into()),
main_language: Some("en".into()),
default_author: None,
max_posts_per_page: 50,
blogmark_category: None,
pico_theme: None,
semantic_similarity_enabled: false,
blog_languages: vec!["en".into(), "de".into()],
}
}
#[test]
fn canonical_post_paths_follow_language_prefix_rule() {
let post = make_post();
assert_eq!(build_canonical_post_path(&post, "en", "en"), "/2024/03/09/hello");
assert_eq!(build_canonical_post_path(&post, "de", "en"), "/de/2024/03/09/hello");
}
#[test]
fn starter_single_post_renderer_uses_canonical_route_and_language_links() {
let post = make_post();
let metadata = make_metadata();
let rendered = render_starter_single_post_page(&post, "Body with [link](/posts/hello)", &metadata, "en").unwrap();
assert_eq!(rendered.relative_path, "2024/03/09/hello/index.html");
assert!(rendered.html.contains("https://example.com/2024/03/09/hello"));
assert!(rendered.html.contains("https://example.com/de/2024/03/09/hello"));
assert!(rendered.html.contains("href=\"/2024/03/09/hello\""));
}
#[test]
fn starter_list_renderer_groups_posts_and_uses_language_specific_index_path() {
let metadata = make_metadata();
let first = make_post();
let mut second = make_post();
second.id = "post-2".into();
second.slug = "next".into();
second.title = "Next".into();
second.published_at = Some(1_710_086_400_000);
second.created_at = 1_710_086_400_000;
let rendered = render_starter_list_page(
&[(first, "First body".into()), (second, "Second body".into())],
&metadata,
"de",
)
.unwrap();
assert_eq!(rendered.relative_path, "de/index.html");
assert!(rendered.html.contains("archive-day-group"));
assert!(rendered.html.contains("09.03.2024"));
assert!(rendered.html.contains("10.03.2024"));
assert!(rendered.html.contains("href=\"/de/2024/03/10/next\""));
}
}

View File

@@ -0,0 +1,75 @@
use std::collections::HashMap;
use crate::model::{Post, Tag, Template, TemplateKind};
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct RenderCategorySettings {
pub post_template_slug: Option<String>,
}
#[derive(Debug, Clone)]
pub struct RenderTemplateLookup<'a> {
pub post: &'a Post,
pub templates: &'a [Template],
pub tags: &'a [Tag],
pub category_settings: &'a HashMap<String, RenderCategorySettings>,
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum TemplateLookupError {
MissingExplicitTemplate(String),
MissingDefaultTemplate,
}
pub fn resolve_post_template<'a>(lookup: RenderTemplateLookup<'a>) -> Result<&'a Template, TemplateLookupError> {
if let Some(explicit_slug) = lookup.post.template_slug.as_deref() {
return lookup
.templates
.iter()
.find(|template| is_enabled_post_template(template, explicit_slug))
.ok_or_else(|| TemplateLookupError::MissingExplicitTemplate(explicit_slug.to_string()));
}
for post_tag in &lookup.post.tags {
if let Some(template_slug) = lookup
.tags
.iter()
.find(|tag| tag.name.eq_ignore_ascii_case(post_tag))
.and_then(|tag| tag.post_template_slug.as_deref())
{
if let Some(template) = lookup
.templates
.iter()
.find(|template| is_enabled_post_template(template, template_slug))
{
return Ok(template);
}
}
}
for category_name in &lookup.post.categories {
if let Some(template_slug) = lookup
.category_settings
.get(category_name)
.and_then(|settings| settings.post_template_slug.as_deref())
{
if let Some(template) = lookup
.templates
.iter()
.find(|template| is_enabled_post_template(template, template_slug))
{
return Ok(template);
}
}
}
lookup
.templates
.iter()
.find(|template| is_enabled_post_template(template, "post"))
.ok_or(TemplateLookupError::MissingDefaultTemplate)
}
fn is_enabled_post_template(template: &Template, slug: &str) -> bool {
template.enabled && template.kind == TemplateKind::Post && template.slug == slug
}