/auto (#16696)

Add `/auto` behind a feature flag that's disabled for now, even for
staff.

We've decided on a different design for context inference, but there are
parts of /auto that will be useful for that, so we want them in the code
base even if they're unused for now.

Release Notes:

- N/A

---------

Co-authored-by: Antonio Scandurra <me@as-cii.com>
Co-authored-by: Marshall Bowers <elliott.codes@gmail.com>

crates/collab/src/db/queries/projects.rs

@@ -728,6 +728,11 @@ impl Database {
                         is_ignored: db_entry.is_ignored,
                         is_external: db_entry.is_external,
                         git_status: db_entry.git_status.map(|status| status as i32),
+                        // This is only used in the summarization backlog, so if it's None,
+                        // that just means we won't be able to detect when to resummarize
+                        // based on total number of backlogged bytes - instead, we'd go
+                        // on number of files only. That shouldn't be a huge deal in practice.
+                        size: None,
                         is_fifo: db_entry.is_fifo,
                     });
                 }

crates/collab/src/db/queries/rooms.rs

@@ -663,6 +663,11 @@ impl Database {
                             is_ignored: db_entry.is_ignored,
                             is_external: db_entry.is_external,
                             git_status: db_entry.git_status.map(|status| status as i32),
+                            // This is only used in the summarization backlog, so if it's None,
+                            // that just means we won't be able to detect when to resummarize
+                            // based on total number of backlogged bytes - instead, we'd go
+                            // on number of files only. That shouldn't be a huge deal in practice.
+                            size: None,
                             is_fifo: db_entry.is_fifo,
                         });
                     }

crates/collab/src/lib.rs

@@ -170,8 +170,8 @@ pub struct Config {
     pub anthropic_api_key: Option<Arc<str>>,
     pub anthropic_staff_api_key: Option<Arc<str>>,
     pub llm_closed_beta_model_name: Option<Arc<str>>,
-    pub qwen2_7b_api_key: Option<Arc<str>>,
-    pub qwen2_7b_api_url: Option<Arc<str>>,
+    pub runpod_api_key: Option<Arc<str>>,
+    pub runpod_api_summary_url: Option<Arc<str>>,
     pub zed_client_checksum_seed: Option<String>,
     pub slack_panics_webhook: Option<String>,
     pub auto_join_channel_id: Option<ChannelId>,
@@ -235,8 +235,8 @@ impl Config {
             stripe_api_key: None,
             stripe_price_id: None,
             supermaven_admin_api_key: None,
-            qwen2_7b_api_key: None,
-            qwen2_7b_api_url: None,
+            runpod_api_key: None,
+            runpod_api_summary_url: None,
             user_backfiller_github_access_token: None,
         }
     }

crates/collab/src/llm.rs

@@ -402,12 +402,12 @@ async fn perform_completion(
         LanguageModelProvider::Zed => {
             let api_key = state
                 .config
-                .qwen2_7b_api_key
+                .runpod_api_key
                 .as_ref()
                 .context("no Qwen2-7B API key configured on the server")?;
             let api_url = state
                 .config
-                .qwen2_7b_api_url
+                .runpod_api_summary_url
                 .as_ref()
                 .context("no Qwen2-7B URL configured on the server")?;
             let chunks = open_ai::stream_completion(

crates/collab/src/llm/db/queries/providers.rs

@@ -1,5 +1,5 @@
 use super::*;
-use sea_orm::QueryOrder;
+use sea_orm::{sea_query::OnConflict, QueryOrder};
 use std::str::FromStr;
 use strum::IntoEnumIterator as _;
 
@@ -99,6 +99,17 @@ impl LlmDatabase {
                     ..Default::default()
                 }
             }))
+            .on_conflict(
+                OnConflict::columns([model::Column::ProviderId, model::Column::Name])
+                    .update_columns([
+                        model::Column::MaxRequestsPerMinute,
+                        model::Column::MaxTokensPerMinute,
+                        model::Column::MaxTokensPerDay,
+                        model::Column::PricePerMillionInputTokens,
+                        model::Column::PricePerMillionOutputTokens,
+                    ])
+                    .to_owned(),
+            )
             .exec_without_returning(&*tx)
             .await?;
             Ok(())

crates/collab/src/llm/db/seed.rs

@@ -40,6 +40,15 @@ pub async fn seed_database(_config: &Config, db: &mut LlmDatabase, _force: bool)
             price_per_million_input_tokens: 25,   // $0.25/MTok
             price_per_million_output_tokens: 125, // $1.25/MTok
         },
+        ModelParams {
+            provider: LanguageModelProvider::Zed,
+            name: "Qwen/Qwen2-7B-Instruct".into(),
+            max_requests_per_minute: 5,
+            max_tokens_per_minute: 25_000, // These are arbitrary limits we've set to cap costs; we control this number
+            max_tokens_per_day: 300_000,
+            price_per_million_input_tokens: 25,
+            price_per_million_output_tokens: 125,
+        },
     ])
     .await
 }

crates/collab/src/tests/test_server.rs

@@ -679,8 +679,8 @@ impl TestServer {
                 stripe_api_key: None,
                 stripe_price_id: None,
                 supermaven_admin_api_key: None,
-                qwen2_7b_api_key: None,
-                qwen2_7b_api_url: None,
+                runpod_api_key: None,
+                runpod_api_summary_url: None,
                 user_backfiller_github_access_token: None,
             },
         })