GameForge Content

AI-generated NPC dialogue, item flavour text, quests, and lore — built by game devs, for game devs.

GameForge Content is a generation API for game writing. Send a structured description of a character, item, quest, or world subject, and get back ready-to-use, in-character text shaped to your genre, tone, and length constraints. Every endpoint is backed by a DeepSeek LLM and returns clean, parseable JSON — no prompt engineering required on your side.

Endpoints

All endpoints are POST, accept a JSON body, and return JSON.

Shared enums: genre accepts high_fantasy, dark_fantasy, sci_fi, western, horror, cozy, cyberpunk. Item type accepts weapon, armor, consumable, quest_item, tool, accessory; item rarity accepts common, uncommon, rare, epic, legendary. Quest type accepts fetch, escort, kill, explore, deliver, investigate, defend. Lore subject type accepts faction, location, historical_event, character_backstory, religion, artifact.

Why this API

• Structured in, structured out. You send typed JSON fields, not raw prompts, and always get back the same predictable response shape — ready to drop straight into a dialogue system, tooltip, or quest log.
• Built for game context. Genre, tone, faction, player relationship, difficulty, narrator voice and more are first-class inputs, so output stays consistent with your world instead of generic AI prose.
• Real generation constraints. Honor hard limits like max_words per dialogue line, max_characters per item description, and target word_count for lore — and the item and lore responses report the actual character_count / word_count back to you.
• Emotion-tagged dialogue. NPC lines can carry an emotion tag (neutral, happy, angry, fearful, sad, curious, impatient) so you can drive portraits, voice direction, or animation cues.
• Response caching built in. Identical requests are served from a Redis cache, so repeated calls are fast and free of duplicate generation; the cached flag in every response tells you whether the text was freshly generated.

Typical use cases

• Populate background NPCs with distinct, in-character barks and conversation lines instead of hand-writing every villager.
• Auto-generate item tooltips and inventory flavour text from the stats and rarity your loot system already produces.
• Spin up side quests on demand — name, description, objectives, and reward text — keyed to player level, faction, and setting.
• Build out worldbuilding: faction histories, location entries, religions, and artifact lore written in a chosen in-world narrator voice.
• Prototype a game's writing tone quickly by sweeping the same subject across genres and styles.

Good to know

• Generation is LLM-backed, so responses take real time (typically a few seconds) and are non-deterministic — the same input can produce different text on a cache miss. Use the cached flag to tell fresh generations from cache hits.
• Required fields matter. /npc/dialogue needs character.name, role, and personality; /item/description needs item.name, type, and rarity; /quest/generate needs quest.type and setting; /lore/worldbuild needs subject.type and name. Everything in context/options is optional and defaults to a high-fantasy preset.
• Length limits are targets, not hard guarantees. max_words, max_characters, and word_count are passed to the model as instructions; the returned character_count / word_count let you verify and re-roll if a result overshoots.
• type and rarity style fields are enums — values outside the lists above are rejected with a validation error.
• Output is always a JSON object parsed from the model, so you can consume it directly without scraping free-form text.