AI Film Recommender for DataLife Engine

AI Film Recommender adds a dedicated page to the site for selecting movies and TV series based on a user's text query. The query can describe mood, genre, desired atmosphere, plot, country of production, release period, or any other combination of conditions.

The module does not import third-party materials or display works not present in the site's database. AI is used to generate a list of recommendations, after which the module searches for matching DLE publications and displays only the found records.

How It Works

1. The user submits a text query via the module's AJAX interface.
2. The query is supplemented with the site's catalog settings and administrator instructions.
3. The selected AI provider returns a structured list: original title, localized titles, and release year.
4. The module matches the received data against DLE publications using configured extra fields and the material title.
5. Found publications are displayed via the standard show.short.php mechanism and the selected TPL template.

Supported AI Providers

• OpenAI (GPT) — separate API key, model selection, web search, forced web search, and search context depth configuration.
• Google Gemini — separate API key, model selection, and web search via Google Search grounding.
• DeepSeek — separate API key and model selection; built-in web search is not used for this provider.

The list of allowed models is stored within the module. If an unknown or outdated model is selected, the default model for the respective provider is used.

Prompt and Catalog Topic Configuration

The administrator sets the context in which the AI should generate recommendations. Ready-made profiles are available:

• movies and series from various countries;
• Korean dramas;
• anime;
• Turkish series;
• Indian cinema;
• custom profile.

For any profile, you can specify a custom catalog description and additional instructions. This allows tailoring to the site's specialization, title language, database composition, and recommendation rules.

Matching Recommendations with the DLE Database

• Ordered title fields. The administrator selects several extra fields and sets their priority. The first field is the original title, followed by localized titles and the publication title as a fallback source.
• Year verification. Multiple year fields can be selected, with strict, preferred, or disabled matching modes.
• Structured AI response. The search considers the original title, alternative titles, and year, not just a single text string.
• Deterministic selection. In case of multiple matches, the priority of the title source, year match, and publication ID are considered.
• XFields compatibility. For older DLE versions, the xfields.txt format is supported; for newer versions, xfields.json and DLEXFields are supported.

Displaying Results

• Results are generated from real site publications.
• You can specify a separate TPL template for cards without modifying the main shortstory.tpl.
• Standard DLE tags and additional fields processed via show.short.php are supported.
• The number of displayed recommendations is configured in the admin panel.
• The module page has separate SEO Title, Description, and Keywords.

Caching and Request Limits

• AI responses are stored in a structured JSON cache, taking into account the provider, model, web search mode, site profile, and other request parameters.
• Different profiles and settings do not share a common cache, even if the user submits the same text.
• Old cache from previous versions is only read in a compatible context and migrated to the new format.
• A separate hourly request limit is set for each user group.
• The counter increments only after a successful result is received and generated.
• The admin panel provides a view of the cache status and allows safe clearing of the module cache.

Proxy and Network Requests

For AI API requests, a common proxy can be enabled. HTTP, SOCKS5, and SOCKS5H are supported, along with username and password authentication. SOCKS5H performs domain name resolution through the proxy.

URL and SEO Routes

• The module is always accessible via the internal URL index.php?do=ai-film-recommender.
• You can enable a separate SEO-friendly prefix for the page, e.g., /ai-films/.
• On DLE 19.0 and newer, module rules are synchronized with rules.json.
• During synchronization, conflicts are checked, a backup is created, and only rules belonging to the module are modified.
• For older DLE versions, the panel displays rules for manual web server configuration.

Admin Panel

The control panel is designed using standard DLE components and is divided into three parts:

• Settings — providers, models, web search, proxy, site profile, field mapping, result template, limits, and SEO;
• Cache Management — information about cache files, verification, and clearing;
• URL and SEO Routes — configuration and synchronization of SEO-friendly URLs.

API keys and proxy passwords are not displayed after saving. Administrative actions check access rights and the user_hash token.

Compatibility and Requirements

• DataLife Engine: 14.x–20.0.
• PHP: 7.4 and newer.
• PHP Extensions: cURL, JSON, mbstring.
• AI API: Requires a key from at least one supported provider.
• Material Database: For accurate search, at least one real additional field for the original title must be configured; a year field is recommended but not required.

Operating Features

• The quality of the result depends on the composition of the site's database, the completeness of titles and the year of release, the selected model, and the user's query text.
• Web search is only used by providers that support the corresponding mode.
• The module outputs only those recommendations for which a match is found among the site's publications.
• The editable configuration file is created after saving the settings and is not included in the installation archive, so updating the module does not overwrite the site's settings.

Video Preview