Amortize LLM research cost across generated articles (v0.5.2)

Research LLM call runs once per pipeline but was logged with no run_id
or post_id, making it invisible in per-article cost breakdowns. Now the
pipeline's cost tracker flows through Source_Collector to the research
provider, tagging costs with the run_id. After all articles complete,
amortize_research_costs() divides the cost evenly and inserts one row
per article. link_generation_logs() excludes llm_research stage so the
first article doesn't grab the full cost.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
peptiderepo
2026-04-20 00:44:33 +07:00
parent 327cdc08d9
commit 264fcf0f84
6 changed files with 153 additions and 7 deletions

View File

@@ -75,6 +75,8 @@ class PRAutoBlogger_Pipeline_Runner {
if ( ! empty( $ideas ) ) {
$this->queue_remaining( $ideas, $cost_tracker->get_run_id(), $result );
} else {
// Single-article run — amortize research costs now (1 article = full cost).
PRAutoBlogger_Post_Assembler::amortize_research_costs( $cost_tracker->get_run_id() );
$this->log_summary( $result );
}
@@ -134,6 +136,8 @@ class PRAutoBlogger_Pipeline_Runner {
$this->update_queue_status( $queue );
$this->schedule_next();
} else {
// All articles done — amortize shared research costs across them.
PRAutoBlogger_Post_Assembler::amortize_research_costs( $queue['run_id'] );
$this->write_final_status( $queue['results'] );
delete_option( self::QUEUE_KEY );
PRAutoBlogger_Generation_Lock::release();
@@ -153,7 +157,9 @@ class PRAutoBlogger_Pipeline_Runner {
$target = absint( get_option( 'prautoblogger_daily_article_target', 1 ) );
$this->broadcast_stage( __( 'Collecting sources from Reddit…', 'prautoblogger' ) );
( new PRAutoBlogger_Source_Collector() )->collect_from_all_sources();
( new PRAutoBlogger_Source_Collector() )
->set_cost_tracker( $cost_tracker )
->collect_from_all_sources();
$this->broadcast_stage( __( 'Analyzing topics and scoring…', 'prautoblogger' ) );
$llm = new PRAutoBlogger_OpenRouter_Provider();

View File

@@ -58,10 +58,13 @@ class PRAutoBlogger_Post_Assembler {
$table = $wpdb->prefix . 'prautoblogger_generation_log';
if ( null !== $run_id && '' !== $run_id ) {
// Exclude 'llm_research' stage — those shared costs are amortized
// across all articles after the full pipeline completes, not linked
// to a single post. See amortize_research_costs().
// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery
$wpdb->query(
$wpdb->prepare(
"UPDATE {$table} SET post_id = %d WHERE post_id IS NULL AND run_id = %s",
"UPDATE {$table} SET post_id = %d WHERE post_id IS NULL AND run_id = %s AND stage != 'llm_research'",
$post_id,
$run_id
)
@@ -79,6 +82,97 @@ class PRAutoBlogger_Post_Assembler {
}
}
/**
* Amortize shared LLM research costs across all articles in a pipeline run.
*
* The research LLM call runs once per pipeline execution before any articles
* are generated. Its cost is logged with the run_id but no post_id. After all
* articles finish, this method divides the research cost evenly and inserts
* one 'llm_research' row per article so the cost breakdown popover includes
* the amortized research overhead.
*
* Side effects: database SELECT, INSERT, DELETE on generation_log.
*
* @param string|null $run_id Pipeline run UUID. Null or empty is a no-op.
*/
public static function amortize_research_costs( ?string $run_id ): void {
if ( null === $run_id || '' === $run_id ) {
return;
}
global $wpdb;
$table = $wpdb->prefix . 'prautoblogger_generation_log';
// Find the unlinked research cost row for this run.
// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery
$research_row = $wpdb->get_row(
$wpdb->prepare(
"SELECT id, estimated_cost, provider, model, prompt_tokens, completion_tokens
FROM {$table}
WHERE run_id = %s AND stage = 'llm_research' AND post_id IS NULL
LIMIT 1",
$run_id
)
);
if ( ! $research_row || (float) $research_row->estimated_cost <= 0.0 ) {
return; // No research cost to amortize (LLM research not enabled or free).
}
// Count distinct articles produced in this run.
// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery
$post_ids = $wpdb->get_col(
$wpdb->prepare(
"SELECT DISTINCT post_id FROM {$table} WHERE run_id = %s AND post_id IS NOT NULL",
$run_id
)
);
$post_count = count( $post_ids );
if ( $post_count < 1 ) {
return; // No articles produced — nothing to amortize into.
}
$total_cost = (float) $research_row->estimated_cost;
$amortized_cost = $total_cost / $post_count;
$total_prompt = (int) $research_row->prompt_tokens;
$total_completion = (int) $research_row->completion_tokens;
$amortized_prompt = (int) round( $total_prompt / $post_count );
$amortized_compl = (int) round( $total_completion / $post_count );
// Insert one amortized research row per article.
foreach ( $post_ids as $post_id ) {
// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery
$wpdb->insert( $table, [
'post_id' => (int) $post_id,
'run_id' => $run_id,
'stage' => 'llm_research',
'provider' => $research_row->provider,
'model' => $research_row->model,
'prompt_tokens' => $amortized_prompt,
'completion_tokens' => $amortized_compl,
'estimated_cost' => $amortized_cost,
'response_status' => 'success',
'error_message' => '',
'created_at' => current_time( 'mysql' ),
] );
}
// Remove the original unlinked row — it's been replaced by per-article rows.
// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery
$wpdb->delete( $table, [ 'id' => (int) $research_row->id ] );
PRAutoBlogger_Logger::instance()->info(
sprintf(
'Amortized LLM research cost $%.4f across %d articles ($%.4f each).',
$total_cost,
$post_count,
$amortized_cost
),
'cost-tracker'
);
}
/**
* Strip common LLM output artifacts from generated content.
* Removes markdown fences, HTML document wrappers, stray head-level tags.

View File

@@ -20,6 +20,24 @@ declare(strict_types=1);
*/
class PRAutoBlogger_Source_Collector {
/**
* Pipeline cost tracker — passed to providers that need run_id tagging.
* Null when called outside the pipeline (e.g., manual test collection).
*/
private ?PRAutoBlogger_Cost_Tracker $cost_tracker = null;
/**
* Inject the pipeline's cost tracker so source providers can log costs
* with the correct run_id. Without this, provider costs are orphaned.
*
* @param PRAutoBlogger_Cost_Tracker $cost_tracker Pipeline cost tracker with run_id set.
* @return $this
*/
public function set_cost_tracker( PRAutoBlogger_Cost_Tracker $cost_tracker ): self {
$this->cost_tracker = $cost_tracker;
return $this;
}
/**
* Collect data from all enabled source providers and store in database.
*
@@ -219,10 +237,16 @@ class PRAutoBlogger_Source_Collector {
}
if ( 'llm_research' === $source_type ) {
return [
$config = [
'model' => get_option( 'prautoblogger_research_model', PRAUTOBLOGGER_DEFAULT_ANALYSIS_MODEL ),
'prompt' => get_option( 'prautoblogger_research_prompt', '' ),
];
// Pass the pipeline's cost tracker so research costs get tagged
// with the run_id for post-pipeline amortization across articles.
if ( null !== $this->cost_tracker ) {
$config['cost_tracker'] = $this->cost_tracker;
}
return $config;
}
return [];

View File

@@ -34,7 +34,7 @@ class PRAutoBlogger_LLM_Research_Provider implements PRAutoBlogger_Source_Provid
*
* Side effects: LLM API call via OpenRouter, cost logging.
*
* @param array{prompt?: string, model?: string} $config Research-specific configuration.
* @param array{prompt?: string, model?: string, cost_tracker?: PRAutoBlogger_Cost_Tracker} $config Research-specific configuration.
* @return PRAutoBlogger_Source_Data[] Array of research findings as source data.
* @throws \RuntimeException On API error after retries.
*/
@@ -65,7 +65,9 @@ class PRAutoBlogger_LLM_Research_Provider implements PRAutoBlogger_Source_Provid
]
);
$cost_tracker = new PRAutoBlogger_Cost_Tracker();
// Use the pipeline's cost tracker if provided (for run_id tagging),
// otherwise create a standalone one (backward compat for manual calls).
$cost_tracker = $config['cost_tracker'] ?? new PRAutoBlogger_Cost_Tracker();
$cost_tracker->log_api_call(
null,
'llm_research',