Generation Modes Reference

---

Mode Dispatch Order

When a worker runs, modes are checked in this order: Compose → Blaze → Simple. Only the first populated mode executes.

---

Simple Mode

One AI call returns the full article.

Config fields:

Field	Type	Required	Description
prompt	InputPrompt	Yes	The prompt configuration (see InputPrompt Fields below)
auto_internal_link	boolean	No	Auto-inject internal links
image_gen	object	No	Image generation config

Execution:

1. Inject common placeholders ({topic}, {niche}, {language}, {keyword})
2. Expand scraper macros
3. If prompt starts with {debug}, return injected prompt without calling AI
4. Send single prompt to model
5. Extract meta tags from generated HTML

Best for: Compact one-shot prompts, feed/SILO automation, quick generation.

---

Blaze Mode

Multiple sequential AI calls assemble a structured article.

Config fields:

Field	Type	Required	Description
title_prompt	InputPrompt	Yes	Generates article title
lsi_prompt	InputPrompt	Yes	Generates section outline (LSI list)
detail_prompt	InputPrompt	Yes	Generates each section body
meta_prompt	InputPrompt	No	Generates meta description
opening_prompt	InputPrompt	No	Generates introductory paragraph
closing_prompt	InputPrompt	No	Generates conclusion paragraph
internal_link_prompt	InputPrompt	No	Generates internal link candidates
internal_link_target	string	No	URL for internal link source
auto_internal_link	boolean	No	Auto-inject internal links
image_gen	object	No	Image generation config

Execution order:

1. Generate title
2. Generate outline (LSI section list)
3. Generate internal links (if configured)
4. Generate meta / opening / closing (in parallel, if configured)
5. Generate each section body (one call per section)
6. Inject internal links into selected sections
7. Concatenate: opening + sections + closing (meta stored separately)

Best for: Structured long-form articles, multi-section content, detailed research pieces.

---

Compose Mode

Section-based composition — each section has its own prompt, joined by a glue string.

Config fields:

Field	Type	Required	Description
title_prompt	InputPrompt	No	Generates article title
meta_prompt	InputPrompt	Yes	Generates meta description
contents_prompt	InputPrompt[]	Yes	Array of section prompts
glue_string	string	Yes	String inserted between sections
internal_link_target	string[]	No	URLs for internal link injection
auto_internal_link	boolean	No	Auto-inject internal links
image_gen	object	No	Image generation config

Execution:

• If any contents_prompt contains {chain}: sequential (chained) mode — sections run one at a time in order, each receives prior output as {chain}
• Otherwise: parallel mode — all sections run simultaneously
• Final article = title + sections joined by glue_string

Best for: Flexible section-based composition, chained context (each section builds on previous), reusing modular section prompts.

---

InputPrompt Fields

Each InputPrompt object has:

Field	Type	Description
prompt	string	Prompt text (supports variables and macros)
temperature	float	Creativity/randomness (0.0–2.0)
model	string (optional)	Model override (uses Settings default if blank)
webSearch	boolean (optional)	Enable web search for this prompt
imagegen	boolean (optional)	Generate image for this prompt's output

---

Placeholder Variables

Common placeholders injected before model call:

Variable	Description
{topic}	Article topic / keyword
{keyword}	Same as {topic}
{niche}	Project niche/vertical
{language}	Content language
{topicQueryParam}	URL-encoded topic (for search macros)

Blaze-specific:

• {title} — generated article title
• {lsi}, {section}, {outline} — outline/LSI section list
• {index}, {item} — section index/name
• {internal_links} — internal link candidates
• {subtopic} — subsection topic

Compose-specific:

• {title} — generated article title
• {index} — section index
• {chain} — output from prior section (chained mode only)
• {internalLink} — internal link candidate
• {random5InternalLink} — 5 random internal link candidates

See Prompt Variables Reference for full details.