silas 0.3.0 → 0.3.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 187a0730a1a84463f6207060d446c6960946513c0884d708fc0d16c011493a26
-  data.tar.gz: ed9a09b617c7dde80deaf6fe2c460e254cc96487b43eac52f56c6aa7dedecbaa
+  metadata.gz: a8ece90334c6059bd396179bfae54b62ebd8abc483d72309da718dbb8e147845
+  data.tar.gz: '014359b0b35a36e7b34b796fbb9194f99c49bf39a1901b4b82cb7b94a10c0ea0'
 SHA512:
-  metadata.gz: a38278a7020a9282b1ff9aaac7b18da60bd017a001bcc421c6e316d0033b1a4b468cd86c0d4bcdcd869940f7d9c4b953318e84dd626087bef4fee1ee8ea8b87c
-  data.tar.gz: 6f42e1a81bf935ebdbb39c678e46662ab85b9f45bd8293c0e83f4467e7db922e2102087ab62ea7eb8e8b1b3b1d323eba1a5772e997028b3e5838882bda9cfeb3
+  metadata.gz: 25e3610da0c9a62b6e9f2a0f956e01c989c62eec40bfdee6703b1bd9206821bb6b4a7988726671b57cd53949ae518239a76d76511cc260b3279db9a36b4f38c4
+  data.tar.gz: f891bf52671a8cd2c7fdc9041b036194939995dc4de7dee7bdd8f09cea986593bc526681d161d41add81b16cfe9257d75df7599e09a7fb85d5af52a9b117828d

data/CHANGELOG.md

@@ -1,5 +1,53 @@
 # Changelog
 
+## 0.3.2
+
+- **The `timeout` budget no longer counts time spent parked for approval.**
+  The wall clock restarts when an approval resumes a turn, because any
+  approval slower than `limits.timeout` previously made the approved resume
+  *instantly* re-park on "timeout" — pathological for a gate whose whole point
+  is waiting for a person (found live in the playground: a 3-minute approval
+  against a 120s timeout). Timeout now bounds **active** stretches — hung
+  providers, runaway loops; crash-rescue resumes keep the original clock (the
+  turn was genuinely live), and cost/token budgets stay cumulative because
+  they measure real spend.
+- **Broadcast-rendered trace partials work in host apps.** The inbox's live
+  trace renders through Turbo's broadcast jobs — i.e. the HOST's default
+  renderer — where the engine-scoped `TraceHelper` and bare engine route
+  helpers didn't exist, so **every broadcast render raised and the live trace
+  silently never streamed in real host apps** (the gem's specs stubbed the
+  dispatch seam and never rendered). The helper is now registered host-wide,
+  partials build routes context-free via `silas_engine_path` (engine route
+  set + discovered mount point), and four host-renderer regression specs pin
+  the real path. `relative_time` is renamed `silas_relative_time` (it is now
+  host-visible, and the bare name is exactly what a host app would define).
+  Also new: `examples/playground` gets a customer-facing chat page that
+  renders and live-streams the engine's own trace partials with zero custom
+  streaming code, plus a scripted keyless demo mode (`bin/setup && bin/dev`
+  with no API key).
+
+## 0.3.1
+
+- **Fixed: only the first scenario in a `silas:eval` run was really tested.**
+  `Silas.configure` never invalidated the memoized resolved engine, so a
+  second `configure` in the same process kept serving the first one — every
+  eval scenario after the first silently ran the *first* scenario's script
+  while still reporting pass/fail as though it hadn't. Reconfiguring now
+  re-resolves both the engine and the sandbox. The gem's own specs missed this
+  because the spec helper resets config between examples; a real multi-scenario
+  eval suite — the production path — does not. Found by building
+  `examples/playground`.
+- **Automatic approvals are now recorded.** A graded gate that clears a call
+  (a lambda under its threshold, an `:once` rule already satisfied) sets
+  `approval_state: "approved"` with **no** `approved_by` — so the audit trail
+  distinguishes "gate ran and passed automatically" (which the inbox now
+  renders as *auto-approved by policy*) from "no gate at all", which stays
+  `nil`. Previously both were indistinguishable on a money-moving tool.
+- The generated `ruby_llm.rb` initializer opts into `use_new_acts_as`, silencing
+  RubyLLM's legacy-API deprecation warning on every boot — Silas never uses
+  `acts_as_*`, so the warning was pure noise. Existing initializers are still
+  never touched.
+
 ## 0.3.0
 
 - **`bin/rails silas:doctor`** — every known first-run failure mode as one

data/app/helpers/silas/inbox/trace_helper.rb

@@ -37,11 +37,21 @@ module Silas
         end
       end
 
-      def relative_time(time)
+      # Prefixed: this helper is registered host-wide (broadcast renders need
+      # it), and "relative_time" is exactly the name a host app would define.
+      def silas_relative_time(time)
         return "" unless time
 
         "#{time_ago_in_words(time)} ago"
       end
+
+      # Engine paths that resolve in EVERY render context. The mounted proxy
+      # (`silas.`) leans on the rendering scope's url_options, which Turbo's
+      # bare broadcast renderer doesn't have — so broadcast-rendered partials
+      # build paths from the engine's own route set + the discovered mount.
+      def silas_engine_path(helper, *args)
+        Silas::Engine.routes.url_helpers.public_send(helper, *args, script_name: Silas::Inbox.mount_path)
+      end
     end
   end
 end

data/app/models/silas/tool_invocation.rb

@@ -84,7 +84,13 @@ module Silas
       return if turn.reload.canceled? || turn.failed? # settled turns never zombie-resume
       return if turn.tool_invocations.where(approval_state: "required").exists?
 
-      turn.update!(status: "queued")
+      # Restart the wall clock: `limits.timeout` bounds ACTIVE stretches (hung
+      # providers, runaway loops), not human deliberation. Without this, any
+      # approval that took longer than the timeout made the approved resume
+      # instantly re-park on "timeout" — pathological for a gate whose whole
+      # point is waiting for a person. Cost/token budgets stay cumulative;
+      # they measure real spend.
+      turn.update!(status: "queued", started_at: Time.current)
       AgentLoopJob.perform_later(turn.id)
     end
   end

data/app/views/silas/inbox/invocations/_approval_card.html.erb

@@ -1,10 +1,13 @@
 <div class="approval">
   <h3>Approval needed — <%= invocation.tool_name %></h3>
   <pre><%= pretty_args(invocation.arguments) %></pre>
-  <%= form_with url: approve_inbox_invocation_path(invocation), method: :post, class: "inline" do %>
+  <%# silas_engine_path, not bare helpers: this partial is broadcast-rendered
+      through the HOST's renderer, where engine route helpers don't exist and
+      the mounted proxy has no routing scope to lean on. %>
+  <%= form_with url: silas_engine_path(:approve_inbox_invocation_path, invocation), method: :post, class: "inline" do %>
     <button class="btn approve">Approve</button>
   <% end %>
-  <%= form_with url: decline_inbox_invocation_path(invocation), method: :post, class: "decline-form" do %>
+  <%= form_with url: silas_engine_path(:decline_inbox_invocation_path, invocation), method: :post, class: "decline-form" do %>
     <textarea name="reason" rows="2" placeholder="Reason (optional — sent back to the agent as the tool result)"></textarea>
     <button class="btn decline">Decline</button>
   <% end %>

data/app/views/silas/inbox/invocations/_invocation.html.erb

@@ -3,8 +3,8 @@
     <code><%= invocation.tool_name %></code>
     <%= status_pill(invocation.awaiting_approval? ? "required" : invocation.status) %>
     <%# The audit line — who held the lever, and why it moved. %>
-    <% if invocation.approval_state == "approved" && invocation.approved_by.present? %>
-      <span class="muted">approved by <%= invocation.approved_by %></span>
+    <% if invocation.approval_state == "approved" %>
+      <span class="muted"><%= invocation.approved_by.present? ? "approved by #{invocation.approved_by}" : "auto-approved by policy" %></span>
     <% elsif invocation.approval_state == "declined" %>
       <span class="muted">declined<%= " by #{invocation.approved_by}" if invocation.approved_by.present? %><%= " — “#{invocation.decline_reason}”" if invocation.decline_reason.present? %></span>
     <% elsif invocation.approval_state == "expired" %>

data/app/views/silas/inbox/sessions/index.html.erb

@@ -45,7 +45,7 @@
       <div class="muted">
         <%= pluralize(session.turns.size, "turn") %> ·
         <%= session.channel.presence || "direct" %> ·
-        <%= relative_time(session.updated_at) %>
+        <%= silas_relative_time(session.updated_at) %>
       </div>
     <% end %>
   <% end %>

data/app/views/silas/inbox/sessions/show.html.erb

@@ -11,7 +11,7 @@
     <span class="name"><%= @session.agent_name %></span>
     <span class="muted">#<%= @session.id %></span>
   </div>
-  <div class="muted"><%= @session.channel.presence || "direct" %> · started <%= relative_time(@session.created_at) %></div>
+  <div class="muted"><%= @session.channel.presence || "direct" %> · started <%= silas_relative_time(@session.created_at) %></div>
   <div id="silas-session-<%= @session.id %>-cost"><%= render "silas/inbox/sessions/cost", session: @session %></div>
 </div>
 

data/app/views/silas/inbox/turns/_header.html.erb

@@ -2,7 +2,10 @@
   <span class="turn-input"><%= truncate(turn.input, length: 90) %></span>
   <%= status_pill(turn.status) %>
   <% if turn.active? && !turn.cancel_requested_at %>
-    <%= form_with url: cancel_inbox_turn_path(turn), method: :post, class: "inline" do %>
+    <%# silas_engine_path — this partial is broadcast-rendered through the
+        host renderer on every turn status change; bare engine helpers (and
+        the scope-dependent mounted proxy) 500 there. %>
+    <%= form_with url: silas_engine_path(:cancel_inbox_turn_path, turn), method: :post, class: "inline" do %>
       <button class="btn-cancel" title="Honored at the next step boundary">Cancel</button>
     <% end %>
   <% elsif turn.cancel_requested_at && turn.active? %>
@@ -14,7 +17,7 @@
     <h3>Budget reached — <%= turn.failure_reason %></h3>
     <div class="muted">Parked at zero compute. Raise the limit and the turn resumes
       by replaying completed work from rows — no re-calls, no re-effects.</div>
-    <%= form_with url: raise_budget_inbox_turn_path(turn), method: :post, class: "topup-form" do %>
+    <%= form_with url: silas_engine_path(:raise_budget_inbox_turn_path, turn), method: :post, class: "topup-form" do %>
       <input name="value" type="number" step="any" min="0" required class="topup-input"
              placeholder="<%= turn.failure_reason == "max_cost" ? "new limit in dollars, e.g. 2.50" : "new limit, e.g. 200000" %>">
       <button class="btn approve">Raise &amp; resume</button>

data/lib/generators/silas/install/templates/ruby_llm.rb

@@ -5,6 +5,12 @@
 #
 # Any provider RubyLLM supports works the same way (openai_api_key, etc.).
 RubyLLM.configure do |c|
+  # Silas never uses RubyLLM's acts_as_* ActiveRecord mixins (it owns its own
+  # durable schema), so opt into the new API to silence the legacy deprecation
+  # warning on every boot. Remove this line if THIS app uses acts_as_chat and
+  # hasn't migrated yet — see https://rubyllm.com/upgrading-to-1-7/
+  c.use_new_acts_as = true if c.respond_to?(:use_new_acts_as=)
+
   c.anthropic_api_key = ENV["ANTHROPIC_API_KEY"]
   # The per-request timeout (seconds; RubyLLM default 300). Under streaming
   # this is an idle-between-chunks timeout — the hang protection for a stuck

data/lib/silas/budget.rb

@@ -7,9 +7,11 @@ module Silas
   #
   # Token/cost checks are deterministic (persisted step data). The timeout check
   # reads the wall clock — benign non-determinism: a cap firing later on resume
-  # is correct (the turn genuinely ran too long across the crash). Note the
-  # timeout clock includes time spent parked, so a timeout top-up should be
-  # sized from Time.current - started_at, not from the original limit.
+  # is correct (the turn genuinely ran too long across the crash). The clock
+  # RESTARTS when an approval resumes the turn (resume_turn! resets
+  # started_at): timeout bounds active stretches, never the hours a human
+  # spends deciding. Crash-rescue resumes keep the original clock — the turn
+  # was live the whole time.
   module Budget
     REASONS = %w[max_input_tokens max_cost timeout].freeze
 

data/lib/silas/engine.rb

@@ -43,6 +43,17 @@ module Silas
       Silas::Inbox::DeltaBroadcaster.subscribe!
     end
 
+    # The trace partials render in TWO contexts: the inbox controllers (which
+    # declare this helper) and Turbo's broadcast jobs, which render through the
+    # HOST's default renderer — where an engine-scoped helper doesn't exist and
+    # every broadcast died silently inside Turbo's job. Register it host-wide
+    # so broadcast renders — and host apps embedding the trace partials in
+    # their own pages — resolve it. (Routes inside those partials go through
+    # TraceHelper#silas_engine_path, which needs no routing scope at all.)
+    initializer "silas.trace_helper" do
+      ActiveSupport.on_load(:action_controller_base) { helper Silas::Inbox::TraceHelper }
+    end
+
     # Registry rebuilds on every code reload in development, once in production.
     initializer "silas.registry" do |app|
       next unless app.root.join("app/agent").exist? || app.root.join("app/agents").exist?

data/lib/silas/inbox.rb

@@ -19,5 +19,20 @@ module Silas
     def stream_name(session_id)
       "silas:inbox:session:#{session_id}"
     end
+
+    # Where the host mounted the engine ("/silas" by the installer's
+    # convention), discovered from the app's route set. Used to build paths
+    # that work in EVERY render context — including Turbo's bare broadcast
+    # renderer, which has no routing scope for the mounted proxy to lean on.
+    def mount_path
+      @mount_path ||= begin
+        route = Rails.application.routes.routes.find do |r|
+          r.app.respond_to?(:app) && r.app.app == Silas::Engine
+        end
+        route&.path&.spec.to_s.sub(/\(.*\z/, "").presence || "/silas"
+      end
+    end
+
+    def reset_mount_path! = (@mount_path = nil) # specs remount
   end
 end

data/lib/silas/ledger.rb

@@ -92,8 +92,15 @@ module Silas
           when Hash # {denied: "reason"} — eve's shape
             invocation.update!(status: "failed", result: { "denied" => verdict[:denied] })
             return :done
+          when :approved
+            # A gate ran and CLEARED it (a lambda under its threshold, or an
+            # :once rule already satisfied). Record that, so the audit trail
+            # can tell "gate evaluated, passed automatically" apart from "no
+            # gate at all" (which stays nil). approved_by is left nil — that
+            # absence is what marks it automatic rather than human.
+            invocation.update!(approval_state: "approved")
           end
-          # :not_applicable / :approved / :once-satisfied fall through
+          # :not_applicable falls through ungated
         end
 
         execute!(invocation, tool)

data/lib/silas/version.rb

@@ -1,3 +1,3 @@
 module Silas
-  VERSION = "0.3.0"
+  VERSION = "0.3.2"
 end

data/lib/silas.rb

@@ -51,9 +51,17 @@ module Silas
       @config ||= Configuration.new
     end
 
+    # Reconfiguring re-resolves the engine and sandbox. Without this the
+    # memos below outlive the config that produced them: a second
+    # `Silas.configure { |c| c.engine = ... }` in one process kept serving the
+    # FIRST engine. That silently broke multi-scenario `silas:eval` runs —
+    # every scenario after the first ran the first one's script and still
+    # reported pass/fail as if it hadn't. Re-resolution is cheap (one `.new`).
     def configure
       yield config
       config.validate!
+      @resolved_engine = nil
+      @resolved_sandbox = nil
       config
     end
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: silas
 version: !ruby/object:Gem::Version
-  version: 0.3.0
+  version: 0.3.2
 platform: ruby
 authors:
 - Daniel St Paul
@@ -103,7 +103,7 @@ description: 'An agent framework where your Rails app is the chassis: app/agent/
   convention, durable runs on Active Job Continuations, exactly-once tool execution
   via a transactional ledger, approvals that park at zero compute.'
 email:
-- daniel@zerogravity.co.uk
+- danielstpaul@hotmail.com
 executables: []
 extensions: []
 extra_rdoc_files: []