samagotchi 0.45.0 → 0.46.0

data/lib/samagotchi/tools/delegate_wait.rb

@@ -84,10 +84,11 @@ module Samagotchi
           outcomes << outcome.line
           return with_outcomes(canceled_result(child_id, reports: reports), outcomes) if outcome.stopped
 
-          # Wait on with the child as it is now: a turn that ends with no
-          # reply after the relay still ends the wait, and the relayed
-          # question isn't reported again.
-          baseline = ReplyWait.baseline_of(Session.load(child_id, state_dir: sd), question_id: wait.question[:id])
+          # Wait on from the child as it was before the relay (read when it
+          # asked), the relayed question no longer reported: a turn that
+          # ends with no reply during the relay's POST or after it still
+          # ends the wait.
+          baseline = ReplyWait.baseline_of(wait.session, question_id: wait.question[:id])
         end
       rescue ArgumentError => e
         "Error: #{e.message}"

data/lib/samagotchi/tools/task_runtime.rb

@@ -66,6 +66,8 @@ module Samagotchi
           err: output_io
         )
         Process.detach(pid)
+        # Read now, compared before each signal: a pid reused later is not this task.
+        pid_started = ProcessGroup.start_time(pid)
 
         now = timestamp
         record = {
@@ -76,6 +78,7 @@ module Samagotchi
           "workspace_root" => Dir.pwd,
           "status" => "running",
           "pid" => pid,
+          "pid_started" => pid_started,
           "created_at" => now,
           "started_at" => now,
           "finished_at" => nil,
@@ -190,7 +193,7 @@ module Samagotchi
 
         pid = refreshed["pid"]
         # Checked again just before the signal: never a group chi didn't start.
-        return [mark_not_chis(refreshed), nil] unless ProcessGroup.leader?(pid)
+        return [mark_not_chis(refreshed), nil] unless chis_group?(refreshed, started: refreshed["pid_started"])
 
         refreshed["stop_requested_by"] = by.to_s
         write_record(refreshed)
@@ -210,13 +213,19 @@ module Samagotchi
         return record unless record["status"] == "running"
 
         pid = record["pid"]
-        return record if ProcessGroup.leader?(pid)
+        return record if chis_group?(record)
         # Gone: its exit code says how it ended.
         return mark_finished_without_exit_code(record) if ProcessGroup.signalable?(pid) && !ProcessGroup.alive?(pid)
 
         mark_not_chis(record)
       end
 
+      # Whether the record's pid still leads its own group. With +started+
+      # (the start time recorded at task_create), also the same process: a
+      # ps per call, so only task_stop asks it, just before the signal;
+      # a refresh (each task_wait poll, task_list) checks the group only.
+      def chis_group?(record, started: nil) = ProcessGroup.leader?(record["pid"], started: started)
+
       # A running record whose pid isn't a group chi started (NOT_CHIS_PROCESS)
       # ends as failed, unless its exit code or a stop on disk says more.
       def mark_not_chis(record)

data/lib/samagotchi/version.rb

@@ -1,7 +1,7 @@
 # frozen_string_literal: true
 
 module Samagotchi
-  VERSION = "0.45.0"
+  VERSION = "0.46.0"
   # Sent as the User-Agent of every HTTP request chi makes.
   USER_AGENT = "chi/#{VERSION}"
 end

data/lib/samagotchi/web/app.rb

@@ -443,7 +443,9 @@ module Samagotchi
             warnings.concat(listing.warnings)
             # the web's own spelling, kept as it was: bare on the default host
             listing.rows.each do |row|
-              models << { name: row.host == default_host ? row.id : "#{row.host}:#{row.id}", host: row.host, id: row.id }
+              model = { name: row.host == default_host ? row.id : "#{row.host}:#{row.id}", host: row.host, id: row.id }
+              model[:configured] = true if row.configured
+              models << model
             end
           end
         rescue StandardError => e
@@ -657,7 +659,8 @@ module Samagotchi
         session_json = session_to_json(session, status: displayed_status(session, snapshot, owner: owner), owner: owner)
         # What the worker's server said it served for that model (after a turn).
         if snapshot
-          session_json = session_json.merge(served_model: snapshot["served_model"], served_model_for: snapshot["served_model_for"])
+          session_json = session_json.merge(served_model: snapshot["served_model"], served_model_for: snapshot["served_model_for"],
+                                            served_expected_by: snapshot["served_expected_by"])
           # The model notes its prompt carries now (a /model there changes them).
           session_json = session_json.merge(prompt_notes: snapshot["prompt_notes"]) if snapshot["prompt_notes"].is_a?(Array)
         end

data/lib/samagotchi/web/public/app.js

@@ -1156,7 +1156,7 @@ function renderSessionView(data) {
   setCapabilities(data.capabilities);
   current.status = s.status;
   current.model = s.model_name;
-  current.served = s.served_model ? [s.served_model, s.served_model_for] : null;
+  current.served = s.served_model ? [s.served_model, s.served_model_for, s.served_expected_by || null] : null;
   current.dir = s.working_directory;
   current.owner = s.owner || null;
   current.timing = normalizeTiming(data.timing);
@@ -2007,7 +2007,7 @@ const streamHandlers = {
   },
   generation_completed: (data) => {
     liveTurn.timing.setNote("");
-    if (data?.served_model) current.served = [data.served_model, data.requested_model];
+    if (data?.served_model) current.served = [data.served_model, data.requested_model, data.served_expected_by || null];
     // The generation's speed and the session's token totals, from the server.
     current.timing.tokens = generationTokens(current.timing.tokens, data);
     if (data?.served_model || data?.tokens) updateInfoBar();
@@ -2212,7 +2212,11 @@ const streamHandlers = {
   llm_context_edited: (data) => { turnDom.notice(llmContextNotice(data)); },
   // The provider is asked again after an error, or the turn waits for
   // plugins' setup: a note on the live timing line until it streams.
-  generation_retrying: (data) => liveTurn.timing.setNote(retryStatusLine(data)),
+  generation_retrying: (data) => {
+    // A dropped stream's step asked again: what it streamed is void.
+    if (data?.restarted) turnDom.generationRestarted(data);
+    liveTurn.timing.setNote(retryStatusLine(data));
+  },
   plugin_init_wait: (data) => liveTurn.timing.setNote(initWaitLine(data)),
   plugin_init_started: (data) => onInitEvent("plugin_init_started", data),
   plugin_init_finished: (data) => onInitEvent("plugin_init_finished", data),

data/lib/samagotchi/web/public/ctx.js

@@ -77,12 +77,24 @@ export function speedText(tps, source) {
   return `${source === "estimate" ? "~" : ""}${number} tok/s`;
 }
 
-// A cost in USD: "$0.42"; under a cent, four decimals ("$0.0012"); "" for
-// none or zero. As Formatting#cost_text.
-export function costText(cost) {
+// A cost in USD: "$0.42"; under a cent, four decimals ("$0.0012"); "~$0.12"
+// for an estimate (hosts.<name>.models prices); "" for none or zero. As
+// Formatting#cost_text.
+export function costText(cost, { estimate = false } = {}) {
   const value = num(cost);
   if (value === null || value <= 0) return "";
-  return value >= 0.01 ? `$${value.toFixed(2)}` : `$${value.toFixed(4)}`;
+  return `${estimate ? "~" : ""}${value >= 0.01 ? `$${value.toFixed(2)}` : `$${value.toFixed(4)}`}`;
+}
+
+// The ctx tooltip's cost: the reported cost as is; with an estimate (an
+// older worker or saved file has none) the total with "~" and both parts.
+function costLine(tokens) {
+  const reported = num(tokens.cost_sum) || 0;
+  const estimated = num(tokens.cost_estimate_sum) || 0;
+  if (estimated <= 0) return reported > 0 ? `cost: ${costText(reported)}` : "";
+  const total = costText(reported + estimated, { estimate: true });
+  const parts = reported > 0 ? `${costText(reported)} reported, ${costText(estimated, { estimate: true })} estimated` : "estimated";
+  return `cost: ${total} (${parts})`;
 }
 
 function count(n) {
@@ -142,7 +154,8 @@ export function ctxWindowText(window) {
 
 // The ctx tooltip (info bar and card): +title+, then the session's tokens
 // (in summed over every request, cached with its share, out with the
-// reasoning part) and its cost, when it has any, and the memory indexes
+// reasoning part) and its cost, when it has any (an estimated part with
+// "~", costLine), and the memory indexes
 // its prompt holds (+memoryIndex+, the snapshot's memory_index); with
 // +window+ (ctxWindowText's) the window line under the title. A
 // delegated session keeps its own metrics, so these are this session's only.
@@ -160,8 +173,8 @@ export function tokensTipText(tokens, title = "Context used after the last turn"
   if (cached > 0 && prompt > 0) parts.push(`cached ${count(cached)} (${Math.round((cached * 100) / prompt)}%)`);
   parts.push(`out ${count(completion)}${reasoning > 0 ? ` (reasoning ${count(reasoning)})` : ""}`);
   const lines = [title, `tokens: ${parts.join(" · ")}`];
-  const cost = costText(tokens.cost_sum);
-  if (cost) lines.push(`cost: ${cost}`);
+  const cost = costLine(tokens);
+  if (cost) lines.push(cost);
   lines.push("this session only, all requests");
   return `${lines.join("\n")}${memoryLine}`;
 }

data/lib/samagotchi/web/public/format.js

@@ -255,8 +255,13 @@ export function servedModelDiffers(asked, served) {
 }
 
 // The session's model for the info bar: the served model with a marker when
-// the server serves another one than +servedFor+ (the name asked).
-export function modelLabel(model, served, servedFor) {
+// the server serves another one than +servedFor+ (the name asked), unless
+// +expectedBy+ names the host whose hosts.<name>.models served: lists it.
+export function modelLabel(model, served, servedFor, expectedBy = null) {
+  if (expectedBy && servedModelDiffers(servedFor, served)) {
+    const name = model || "";
+    return { text: name.slice(0, 40), title: `${name}; served: ${served} (expected: hosts.${expectedBy}.models)`, mismatch: false };
+  }
   if (servedModelDiffers(servedFor, served)) {
     return { text: `${served.slice(0, 40)} ⚠`, title: `served: ${served}; asked for ${model}`, mismatch: true };
   }

data/lib/samagotchi/web/public/model_pick.js

@@ -20,7 +20,8 @@ export const RECENT_KEY = "chi_model_recent";
 // The payload's models as rows: {name, host, id, isDefaultHost}, and
 // sampling (the configured request parameters as /model words them) when
 // the model has some, and llm_context (the LLM context it starts under,
-// LLMContextStrategy::Explained#summary) when the server sent one. The
+// LLMContextStrategy::Explained#summary) when the server sent one, and
+// configured when the host declares the id (hosts.<name>.models). The
 // default host's names are bare (name == id); the server's own default,
 // unshifted when no host lists it, has no host and goes with them, shown
 // by its name.
@@ -34,15 +35,18 @@ export function modelRows(payload) {
       const id = host ? String(m.id || name) : name;
       const row = { name, host, id, isDefaultHost: !host || name === id };
       if (m.sampling) row.sampling = String(m.sampling);
+      if (m.configured === true) row.configured = true;
       if (m.llm_context && typeof m.llm_context === "object") row.llm_context = m.llm_context;
       return row;
     });
 }
 
-const byId = (a, b) => a.id.localeCompare(b.id, "en", { sensitivity: "base" });
+const byId = (a, b) => (b.configured === true) - (a.configured === true) ||
+  a.id.localeCompare(b.id, "en", { sensitivity: "base" });
 
 // The rows under their hosts, [{host, isDefault, rows}]: the default host
-// first, the others in the server's order, ids A-Z inside a host.
+// first, the others in the server's order; inside a host the configured
+// ids first, each part A-Z.
 export function groupRows(rows) {
   const groups = [];
   const find = (key) => groups.find((g) => g.key === key);
@@ -61,6 +65,22 @@ export function groupRows(rows) {
   }));
 }
 
+// The word after a row's id: "default" for the default model, else
+// "config" for an id the host declares (hosts.<name>.models), else "".
+export function rowNote(row, defaultName) {
+  if (row.name === defaultName) return "default";
+  return row.configured ? "config" : "";
+}
+
+// A row's tooltip: its configured sampling, and where a configured id
+// comes from; "" when neither.
+export function rowTitle(row) {
+  const parts = [];
+  if (row.sampling) parts.push(`sampling: ${row.sampling}`);
+  if (row.configured) parts.push(`served by config (hosts.${row.host}.models)`);
+  return parts.join("; ");
+}
+
 const SEPARATORS = "/-:._ ";
 const atSegmentStart = (text, i) => i === 0 || SEPARATORS.includes(text[i - 1]);
 

data/lib/samagotchi/web/public/model_picker.js

@@ -12,7 +12,7 @@
 import { escapeHtml } from "./format.js";
 import { moveSelection } from "./command_complete.js";
 import {
-  MODEL_KEY, RECENT_KEY, groupRows, matchModels, modelRows, pickModel, pushRecent, recentRows,
+  MODEL_KEY, RECENT_KEY, groupRows, matchModels, modelRows, pickModel, pushRecent, recentRows, rowNote, rowTitle,
 } from "./model_pick.js";
 
 const PAGE = 8;
@@ -57,9 +57,11 @@ export function createModelPicker({ button, panel, input, list, composer, center
 
   function rowHtml(row, i, { hostMarks = [], idMarks = [], showHost = !row.isDefaultHost } = {}) {
     const host = showHost ? `<span class="model-host">${markedHtml(row.host, hostMarks)}</span><span class="model-dot"> · </span>` : "";
-    const note = row.name === defaultName ? `<span class="model-note">default</span>` : "";
+    const noteText = rowNote(row, defaultName);
+    const note = noteText ? `<span class="model-note">${noteText}</span>` : "";
     const cls = `model-option${i === active ? " active" : ""}${row.name === chosen ? " chosen" : ""}`;
-    const title = row.sampling ? ` title="${escapeHtml(`sampling: ${row.sampling}`)}"` : "";
+    const titleText = rowTitle(row);
+    const title = titleText ? ` title="${escapeHtml(titleText)}"` : "";
     return `<div class="${cls}" role="option" id="model-option-${i}" aria-selected="${i === active}" data-index="${i}"${title}>` +
       `<span class="model-id">${host}${markedHtml(row.id, idMarks)}</span>${note}</div>`;
   }

data/lib/samagotchi/web/public/stage_view.js

@@ -560,6 +560,7 @@ export function createStageView({ historyEl, appendToHistory, isNearBottom: hist
     },
     generationStarted: (data) => view.generationStarted(data),
     chunk: (data) => view.chunk(data),
+    generationRestarted: (data) => view.generationRestarted(data),
     generationCompleted: () => view.generationCompleted(),
     toolStarted: (data) => view.toolStarted(data),
     toolCompleted: (data) => view.toolCompleted(data),

data/lib/samagotchi/web/public/turn_events.js

@@ -209,8 +209,9 @@ export function noticeLine(entry = {}) {
 
 // The live turn's status while the provider is asked again after an error
 // (generation_retrying): "↻ retrying (503) in 4 s, 2/5" (retry 2 of 5).
-export function retryStatusLine({ attempt, max_retries, next_delay, status, error_class } = {}) {
-  const why = status || String(error_class || "").split("::").pop();
+export function retryStatusLine({ attempt, max_retries, next_delay, status, error_class, restarted } = {}) {
+  // restarted: a dropped stream's step asked again (the chat loop's).
+  const why = restarted ? "stream dropped" : status || String(error_class || "").split("::").pop();
   const delay = Number(next_delay);
   const when = Number.isFinite(delay) ? ` in ${delay < 1 ? delay.toFixed(1) : Math.round(delay)} s` : "";
   const count = attempt ? `, ${attempt}${max_retries ? `/${max_retries}` : ""}` : "";

data/lib/samagotchi/web/public/turn_model.js

@@ -12,6 +12,8 @@
 //   a chunk with no iteration goes to the current gen, unless that one is
 //   textDone or already has tool rows (an older server's replay), which
 //   opens a new gen.
+// - generation_retrying with restarted (a dropped stream's step asked
+//   again) empties its gen's thinking and text.
 // - generation_completed marks the current gen textDone; its tool rows
 //   follow it (the kernel emits them after).
 // - tool events go to the gen of their iteration, else to the current gen,
@@ -105,6 +107,20 @@ export function chunk(turn, { text = "", thinking = "", iteration = null } = {})
   return r;
 }
 
+// A step asked again after its stream dropped (generation_retrying with
+// restarted): the open gen of +iteration+ (else the current one) loses what
+// it streamed; null when there is none.
+export function generationRestarted(turn, iteration = null) {
+  const current = currentGen(turn);
+  const gen = iteration == null ? current
+    : [...turn.gens].reverse().find((g) => !g.closed && g.iteration === iteration) ?? null;
+  if (!gen) return null;
+  gen.thinking = "";
+  gen.text = "";
+  gen.textDone = false;
+  return { gen, opened: false, closed: null };
+}
+
 export function generationCompleted(turn) {
   const gen = currentGen(turn);
   if (gen) gen.textDone = true;

data/lib/samagotchi/web/public/turn_view.js

@@ -615,6 +615,25 @@ export function createTurnView({
       apply(r);
       scheduleFlush(r.gen);
     },
+    // A dropped stream's step asked again (generation_retrying with
+    // restarted): its gen's thinking and narration so far go, the step
+    // shows "…" until the new stream's text comes.
+    generationRestarted(data) {
+      if (!turn) return;
+      const r = model.generationRestarted(turn, data?.iteration ?? null);
+      const entry = r && genEls.get(r.gen);
+      if (!entry) return;
+      dirty.delete(r.gen);
+      if (entry.thinkingWrap) {
+        if (entry.tickerTimer) { clearTimeout(entry.tickerTimer); entry.tickerTimer = null; }
+        entry.ticker.close();
+        entry.thinkingWrap.remove();
+      }
+      Object.assign(entry, { thinking: null, thinkingWrap: null, thinkingLine: null, ticker: null, thinkingWritten: 0,
+        box: null, feed: createSentenceFeed() });
+      entry.text.textContent = "…";
+      onChange();
+    },
     generationCompleted() {
       if (!turn) return;
       const r = model.generationCompleted(turn);

data/lib/samagotchi/worker.rb

@@ -16,6 +16,7 @@ require_relative "session_manager"
 require_relative "archive_store"
 require_relative "log"
 require_relative "turn_flow"
+require_relative "llm/errors"
 require_relative "continue_offer"
 require_relative "iteration_limit"
 require_relative "session_commands"
@@ -830,7 +831,7 @@ module Samagotchi
       # failed prompt or wake turn that got nowhere, a canceled continue
       # turn). A failed turn whose work stayed, and a reminder or continue
       # turn that failed (kept with its failed note), read them.
-      settle_child_reports(rolled_back: @turn_flow.restores != restores)
+      settle_child_reports(rolled_back: @turn_flow.restores != restores || merges_lost?(error))
       ring_parent_after_turn
     end
 
@@ -843,6 +844,20 @@ module Samagotchi
       rolled_back ? @child_reports.release : @child_reports.commit
     end
 
+    # Whether +error+ ended the turn without its loop's conversation: a
+    # failure the loops didn't mark (LLM::FailedTurn), one past their end
+    # (the Engine's own end of the turn, say). The Engine kept the turn's
+    # start only, so the reports its loop merged are not in the
+    # conversation: their rings stay for the next turn, even when the turn
+    # itself stays (a reminder or continue turn). Not one raised after the
+    # turn ended (the Engine's post-turn work): the turn was kept as
+    # completed, the reports in it.
+    def merges_lost?(error)
+      return false if error.nil? || error.is_a?(LLM::FailedTurn)
+
+      @session.last_turn&.dig("outcome") != "completed"
+    end
+
     # A delegate child's turn that was its parent's (#reports_to_parent?)
     # rings the parent, after the save: the parent reads a settled child.
     # A question it rang about that still waits makes the next turn ring

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: samagotchi
 version: !ruby/object:Gem::Version
-  version: 0.45.0
+  version: 0.46.0
 platform: ruby
 authors:
 - Dmitry Dedov
@@ -136,6 +136,7 @@ files:
 - docs/internals/gemma4-contract.md
 - docs/internals/llm-context-bench.md
 - docs/internals/llm-context-forget.md
+- docs/internals/llm-context-live.md
 - docs/internals/prompt-caching.md
 - docs/internals/tool-guardrails.md
 - docs/memory.md
@@ -285,6 +286,7 @@ files:
 - lib/samagotchi/hooks/loader.rb
 - lib/samagotchi/hooks/registry.rb
 - lib/samagotchi/hooks/stream_watch.rb
+- lib/samagotchi/host_model.rb
 - lib/samagotchi/host_registry.rb
 - lib/samagotchi/idle_client.rb
 - lib/samagotchi/idle_recap.rb
@@ -348,6 +350,7 @@ files:
 - lib/samagotchi/model_match.rb
 - lib/samagotchi/model_notes.rb
 - lib/samagotchi/model_overlay.rb
+- lib/samagotchi/model_price.rb
 - lib/samagotchi/model_profile.rb
 - lib/samagotchi/model_ref.rb
 - lib/samagotchi/models_command.rb