@pi-in-go/pigpen-context-info 0.0.0-stage → 0.1.0

package/CREDITS.md

@@ -0,0 +1,19 @@
+# Credits
+
+`extensions/context-info` is original work by **Michael Kinsy**, moved into
+Pigpen from the author's personal PiG configuration and released under Pigpen's
+MIT license (see [LICENSE](LICENSE)). It is not derived from another project.
+
+Changes made when it was moved:
+
+- Ported to PiG's public Go SDK, including its error-returning host reads and
+  Pi's nullable context usage (tokens and percent are unknown right after
+  compaction).
+- The status footer is **off by default** (it replaces PiG's footer). Turn it on
+  with the `--context-footer` flag or `/context-footer on`.
+- Removed `/prompt-edit`, the session prompt overlay, the deletion of the overlay
+  file at session start, and the subagent cost section of `/cost`. They belonged
+  to the author's own subagent extension, which is not part of this Package.
+- Windows: git is started without a console window.
+- `/prompts` no longer fails on an agent file shorter than four bytes, and reads
+  frontmatter only when the file starts with it.

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Michael Kinsy
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -1,3 +1,48 @@
-# Temporary Holding Version
+# context-info
 
-This version is a temporary placeholder for this package. An operational version to replace this has been submitted for review and is awaiting a staged release.
+A Go extension for PiG that shows what fills the context window and what the
+session costs. It shows nothing until you ask. In the background it counts tool
+calls and reads the session at start and after compaction to calibrate its token
+estimate.
+
+| Command | Shows |
+|---|---|
+| `/context` (or `Ctrl+Shift+I`) | The context window: system prompt (with skills), tool definitions, conversation, cache, model and thinking level, available room. Uses the provider's live token count when there is one and an estimate (self-calibrating characters per token) when there is not, for example right after compaction. |
+| `/tools` | Every tool, grouped by source, active or not, with call counts. |
+| `/cost` | Tokens and cost by kind and by model, and the active model's per-million-token rates. |
+| `/prompts` | The current system prompt. `/prompts agents` lists agent definitions from `<config home>/agents/*.md`; `/prompts <name>` shows one. |
+| `/context-footer [on\|off]` | A two-line status footer: directory and git branch, context percentage, model, thinking level, cost, tokens, tools and calls. |
+
+The footer **replaces PiG's footer**, so it is off by default. Start PiG with
+`--context-footer` to have it from the start, or use `/context-footer on`. Turning it
+off gives PiG's footer back. While it is off, the extension makes no footer or status
+updates.
+
+## Cost figures are estimates
+
+Cost is tokens times per-million-token rates. The rates come from the active model.
+Models billed by subscription carry no price, so for GitHub Copilot models the
+extension falls back to a list-price table in `extension.go` (dated 2025-06-01,
+verify before relying on it). A model that is in neither shows tokens and `$0.00`.
+
+## Install
+
+```sh
+pig install ./components/context-info
+```
+
+The extension builds from source on first use (Go toolchain required) or fuses into a
+Piglet Binary. No network, no Node runtime. It reads the session through the host and
+files under the PiG config home (`agents/`); it writes nothing. The footer starts
+`git` to show the branch, at most once every two seconds.
+`pig package validate ./components/context-info` validates the Package.
+
+## Test
+
+The tests use PiG's public SDK through a fake host. From the Pigpen root:
+
+```sh
+PIG_BIN=/path/to/pig npm run test:go-ports -- -race
+```
+
+MIT. © Michael Kinsy. See [CREDITS.md](CREDITS.md) for what changed when it moved here.

package/extensions/context-info/context_test.go

@@ -0,0 +1,427 @@
+package contextinfo
+
+import (
+	"sync"
+	"testing"
+
+	sdk "github.com/MichaelKinsy/PiG/extensions/sdk"
+)
+
+func TestFmtNum(t *testing.T) {
+	cases := []struct {
+		in   int
+		want string
+	}{
+		{0, "0"},
+		{999, "999"},
+		{1_000, "1.0k"},
+		{1_500, "1.5k"},
+		{10_000, "10.0k"},
+		{999_999, "1000.0k"},
+		{1_000_000, "1.0M"},
+		{2_500_000, "2.5M"},
+	}
+	for _, tc := range cases {
+		got := fmtNum(tc.in)
+		if got != tc.want {
+			t.Errorf("fmtNum(%d) = %q, want %q", tc.in, got, tc.want)
+		}
+	}
+}
+
+func TestFmtCost(t *testing.T) {
+	cases := []struct {
+		in   float64
+		want string
+	}{
+		{0, "$0.00"},
+		{0.005, "$0.0050"},
+		{0.001, "$0.0010"},
+		{0.01, "$0.01"},
+		{0.10, "$0.10"},
+		{1.0, "$1.00"},
+		{12.345, "$12.35"},
+	}
+	for _, tc := range cases {
+		got := fmtCost(tc.in)
+		if got != tc.want {
+			t.Errorf("fmtCost(%f) = %q, want %q", tc.in, got, tc.want)
+		}
+	}
+}
+
+func TestCommaFmt(t *testing.T) {
+	cases := []struct {
+		in   int
+		want string
+	}{
+		{0, "0"},
+		{999, "999"},
+		{1_000, "1,000"},
+		{10_000, "10,000"},
+		{100_000, "100,000"},
+		{1_000_000, "1,000,000"},
+		{1_234_567, "1,234,567"},
+	}
+	for _, tc := range cases {
+		got := commaFmt(tc.in)
+		if got != tc.want {
+			t.Errorf("commaFmt(%d) = %q, want %q", tc.in, got, tc.want)
+		}
+	}
+}
+
+func TestCharsToTokens(t *testing.T) {
+	// Default charsPerToken = 3.7
+	charsPerToken = 3.7
+	cases := []struct {
+		chars int
+		want  int
+	}{
+		{0, 0},
+		{1, 1},     // ceil(1/3.7) = 1
+		{4, 2},     // ceil(4/3.7) = ceil(1.08) = 2
+		{37, 10},   // ceil(37/3.7) = 10
+		{370, 100}, // exact
+	}
+	for _, tc := range cases {
+		got := charsToTokens(tc.chars)
+		if got != tc.want {
+			t.Errorf("charsToTokens(%d) = %d, want %d", tc.chars, got, tc.want)
+		}
+	}
+}
+
+func TestAnalyzeSystemPrompt(t *testing.T) {
+	prompt := "You are helpful.\n<skill name=\"test\">some skill content</skill>\nDo things."
+	a := analyzeSystemPrompt(prompt)
+
+	if a.skillCount != 1 {
+		t.Errorf("skillCount = %d, want 1", a.skillCount)
+	}
+	if a.totalChars != len(prompt) {
+		t.Errorf("totalChars = %d, want %d", a.totalChars, len(prompt))
+	}
+	// skillChars = len(`<skill name="test">some skill content</skill>`) = 45
+	if a.skillChars != 45 {
+		t.Errorf("skillChars = %d, want 45", a.skillChars)
+	}
+	if a.baseChars != len(prompt)-45 {
+		t.Errorf("baseChars = %d, want %d", a.baseChars, len(prompt)-45)
+	}
+}
+
+func TestAnalyzeSystemPrompt_NoSkills(t *testing.T) {
+	prompt := "Simple prompt with no skills."
+	a := analyzeSystemPrompt(prompt)
+
+	if a.skillCount != 0 {
+		t.Errorf("skillCount = %d, want 0", a.skillCount)
+	}
+	if a.baseChars != len(prompt) {
+		t.Errorf("baseChars = %d, want %d", a.baseChars, len(prompt))
+	}
+}
+
+func TestAnalyzeConversation(t *testing.T) {
+	branch := []sdk.BranchEntry{
+		{Type: "message", Role: "user", Content: "hello"},
+		{Type: "message", Role: "assistant", Content: "hi there", Thinking: "let me think",
+			ToolCalls: []sdk.ToolCallInfo{{Name: "bash", Args: `{"command":"ls"}`}}},
+		{Type: "message", Role: "toolResult", ToolName: "bash", ToolCallID: "tc1", Content: "file.go"},
+		{Type: "other"},
+	}
+
+	a := analyzeConversation(branch)
+
+	if a.messageCount != 3 {
+		t.Errorf("messageCount = %d, want 3", a.messageCount)
+	}
+	if a.toolCallCount != 1 {
+		t.Errorf("toolCallCount = %d, want 1", a.toolCallCount)
+	}
+	if a.toolResultCount != 1 {
+		t.Errorf("toolResultCount = %d, want 1", a.toolResultCount)
+	}
+	if a.thinkingChars != len("let me think") {
+		t.Errorf("thinkingChars = %d, want %d", a.thinkingChars, len("let me think"))
+	}
+}
+
+func TestAnalyzeConversation_CompactionUsesFirstKeptEntryID(t *testing.T) {
+	charsPerToken = 1
+	branch := []sdk.BranchEntry{
+		{ID: "old", Type: "message", Role: "user", Content: "oldoldold"},
+		{ID: "kept", Type: "message", Role: "user", Content: "kept"},
+		{ID: "compact", Type: "compaction", Content: "sum", FirstKeptEntryID: "kept"},
+		{ID: "new", Type: "message", Role: "assistant", Content: "new"},
+	}
+
+	a := analyzeConversation(branch)
+	if a.textChars != len("sum")+len("kept")+len("new") {
+		t.Fatalf("textChars = %d, want summary + kept + new", a.textChars)
+	}
+	if a.messageCount != 3 {
+		t.Fatalf("messageCount = %d, want summary + kept + new", a.messageCount)
+	}
+}
+
+func TestGetSessionCostData(t *testing.T) {
+	invalidateCostCache()
+	branch := []sdk.BranchEntry{
+		{Role: "assistant", ModelID: "test-model", Usage: &sdk.UsageInfo{Input: 1000, Output: 500, CacheRead: 200, CacheWrite: 100}},
+		{Role: "assistant", ModelID: "test-model", Usage: &sdk.UsageInfo{Input: 2000, Output: 1000}},
+		{Role: "user"}, // no usage — skipped
+	}
+	model := &sdk.ModelInfo{
+		ID:                  "test-model",
+		InputCostPer1M:      3.0,
+		OutputCostPer1M:     15.0,
+		CacheReadCostPer1M:  0.3,
+		CacheWriteCostPer1M: 3.75,
+	}
+
+	d := getSessionCostData(branch, model)
+
+	if d.input != 3000 {
+		t.Errorf("input = %d, want 3000", d.input)
+	}
+	if d.output != 1500 {
+		t.Errorf("output = %d, want 1500", d.output)
+	}
+	if d.cacheRead != 200 {
+		t.Errorf("cacheRead = %d, want 200", d.cacheRead)
+	}
+	if d.cacheWrite != 100 {
+		t.Errorf("cacheWrite = %d, want 100", d.cacheWrite)
+	}
+	// Cost: 3000*3/1M = 0.009, 1500*15/1M = 0.0225, 200*0.3/1M = 0.00006, 100*3.75/1M = 0.000375
+	wantInput := 0.009
+	if d.cost.input != wantInput {
+		t.Errorf("cost.input = %f, want %f", d.cost.input, wantInput)
+	}
+	if d.cost.total == 0 {
+		t.Error("cost.total should be > 0")
+	}
+}
+
+func TestGetSessionCostData_FallsBackToCurrentModelForOldSessions(t *testing.T) {
+	invalidateCostCache()
+	branch := []sdk.BranchEntry{
+		{Role: "assistant", Usage: &sdk.UsageInfo{Input: 1000, Output: 500, TotalTokens: 1500}},
+	}
+	model := &sdk.ModelInfo{ID: "claude-sonnet-4.5", Provider: "github-copilot"}
+
+	d := getSessionCostData(branch, model)
+	if d.cost.total == 0 {
+		t.Fatal("cost.total = 0, want fallback Copilot pricing for current model")
+	}
+	if len(d.byModel) != 1 || d.byModel[0].modelID != "claude-sonnet-4.5" {
+		t.Fatalf("byModel = %+v", d.byModel)
+	}
+	if d.totalTokens != 1500 {
+		t.Fatalf("totalTokens = %d, want totalTokens from usage", d.totalTokens)
+	}
+}
+
+func TestGetSessionCostData_ProviderQualifiedModelID(t *testing.T) {
+	invalidateCostCache()
+	branch := []sdk.BranchEntry{
+		{Role: "assistant", ModelID: "github-copilot/claude-sonnet-4.5", Usage: &sdk.UsageInfo{Input: 1000, Output: 500}},
+	}
+
+	d := getSessionCostData(branch, nil)
+	if d.cost.total == 0 {
+		t.Fatal("cost.total = 0, want pricing for provider-qualified model id")
+	}
+	if len(d.byModel) != 1 || d.byModel[0].modelID != "claude-sonnet-4.5" {
+		t.Fatalf("byModel = %+v", d.byModel)
+	}
+}
+
+func TestGetSessionCostData_CacheKeyIncludesLeafID(t *testing.T) {
+	invalidateCostCache()
+	model := &sdk.ModelInfo{ID: "claude-sonnet-4.5", Provider: "github-copilot"}
+	first := []sdk.BranchEntry{{ID: "leaf-a", Role: "assistant", ModelID: "claude-sonnet-4.5", Usage: &sdk.UsageInfo{Input: 1000, Output: 100}}}
+	second := []sdk.BranchEntry{{ID: "leaf-b", Role: "assistant", ModelID: "claude-sonnet-4.5", Usage: &sdk.UsageInfo{Input: 2000, Output: 200}}}
+
+	_ = getSessionCostData(first, model)
+	got := getSessionCostData(second, model)
+	if got.input != 2000 || got.output != 200 {
+		t.Fatalf("cached stale cost data: input/output = %d/%d", got.input, got.output)
+	}
+}
+
+func TestGetSessionCostData_NoModel(t *testing.T) {
+	invalidateCostCache()
+	branch := []sdk.BranchEntry{
+		{Role: "assistant", Usage: &sdk.UsageInfo{Input: 1000, Output: 500}},
+	}
+	d := getSessionCostData(branch, nil)
+
+	if d.input != 1000 {
+		t.Errorf("input = %d, want 1000", d.input)
+	}
+	if d.cost.total != 0 {
+		t.Errorf("cost.total = %f, want 0 (no model pricing)", d.cost.total)
+	}
+}
+
+func TestAddAssistantUsageToCostCacheConcurrent(t *testing.T) {
+	invalidateCostCache()
+	model := &sdk.ModelInfo{ID: "claude-sonnet-4.5", Provider: "github-copilot", InputCostPer1M: 3, OutputCostPer1M: 15}
+
+	const workers = 8
+	const perWorker = 200
+	var wg sync.WaitGroup
+	wg.Add(workers * 2)
+	for w := 0; w < workers; w++ {
+		go func() {
+			defer wg.Done()
+			for i := 0; i < perWorker; i++ {
+				addAssistantUsageToCostCache("claude-sonnet-4.5", sdk.UsageInfo{Input: 10, Output: 5, TotalTokens: 15}, model)
+			}
+		}()
+		go func() {
+			defer wg.Done()
+			for i := 0; i < perWorker; i++ {
+				_, _ = cachedSessionCostData()
+			}
+		}()
+	}
+	wg.Wait()
+
+	got, ok := cachedSessionCostData()
+	if !ok {
+		t.Fatal("cost cache missing after concurrent updates")
+	}
+	wantInput := workers * perWorker * 10
+	wantOutput := workers * perWorker * 5
+	if got.input != wantInput || got.output != wantOutput {
+		t.Fatalf("lost updates under concurrency: input=%d (want %d) output=%d (want %d)", got.input, wantInput, got.output, wantOutput)
+	}
+	if len(got.byModel) != 1 || got.byModel[0].msgs != workers*perWorker {
+		t.Fatalf("byModel = %+v, want one bucket with %d messages", got.byModel, workers*perWorker)
+	}
+}
+
+func TestAddAssistantUsageToCostCacheUpdatesWithoutBranchRebuild(t *testing.T) {
+	invalidateCostCache()
+	model := &sdk.ModelInfo{ID: "claude-sonnet-4.5", Provider: "github-copilot"}
+	_ = getSessionCostData([]sdk.BranchEntry{
+		{ID: "leaf-a", Role: "assistant", ModelID: "claude-sonnet-4.5", Usage: &sdk.UsageInfo{Input: 1000, Output: 500, TotalTokens: 1500}},
+	}, model)
+
+	addAssistantUsageToCostCache("claude-sonnet-4.5", sdk.UsageInfo{Input: 2000, Output: 1000, TotalTokens: 3000}, model)
+	got, ok := cachedSessionCostData()
+	if !ok {
+		t.Fatal("cost cache missing after incremental update")
+	}
+	if got.input != 3000 || got.output != 1500 || got.totalTokens != 4500 {
+		t.Fatalf("incremental totals = input:%d output:%d total:%d", got.input, got.output, got.totalTokens)
+	}
+	if len(got.byModel) != 1 || got.byModel[0].msgs != 2 {
+		t.Fatalf("byModel = %+v, want one bucket with two messages", got.byModel)
+	}
+}
+
+func TestAssistantMapFromEventAcceptsNestedAndFlatShapes(t *testing.T) {
+	cases := []map[string]any{
+		{"message": map[string]any{"Assistant": map[string]any{"role": "assistant", "model": "m"}}},
+		{"message": map[string]any{"assistant": map[string]any{"role": "assistant", "model": "m"}}},
+		{"message": map[string]any{"role": "assistant", "model": "m"}},
+	}
+	for _, tc := range cases {
+		got, ok := assistantMapFromEvent(tc)
+		if !ok || got["model"] != "m" {
+			t.Fatalf("assistantMapFromEvent(%#v) = %#v, %v", tc, got, ok)
+		}
+	}
+	if _, ok := assistantMapFromEvent(map[string]any{"message": map[string]any{"User": map[string]any{"role": "user"}}}); ok {
+		t.Fatal("user message parsed as assistant")
+	}
+}
+
+func TestUsageInfoFromMapAcceptsProviderAndSDKNames(t *testing.T) {
+	got := usageInfoFromMap(map[string]any{
+		"input_tokens":  float64(10),
+		"output_tokens": float64(20),
+		"cache_read":    float64(3),
+		"cache_write":   float64(4),
+		"totalTokens":   float64(37),
+	})
+	if got.Input != 10 || got.Output != 20 || got.CacheRead != 3 || got.CacheWrite != 4 || got.TotalTokens != 37 {
+		t.Fatalf("usageInfoFromMap = %+v", got)
+	}
+}
+
+func TestGetSessionCostData_CacheKeyIncludesCurrentModel(t *testing.T) {
+	invalidateCostCache()
+	branch := []sdk.BranchEntry{
+		{ID: "leaf", Role: "assistant", Usage: &sdk.UsageInfo{Input: 1000, Output: 500, TotalTokens: 1500}},
+	}
+	claude := &sdk.ModelInfo{ID: "claude-sonnet-4.5", Provider: "github-copilot"}
+	gptMini := &sdk.ModelInfo{ID: "gpt-5-mini", Provider: "github-copilot"}
+
+	first := getSessionCostData(branch, claude)
+	second := getSessionCostData(branch, gptMini)
+
+	if len(first.byModel) != 1 || first.byModel[0].modelID != "claude-sonnet-4.5" {
+		t.Fatalf("first byModel = %+v", first.byModel)
+	}
+	if len(second.byModel) != 1 || second.byModel[0].modelID != "gpt-5-mini" {
+		t.Fatalf("cache ignored current model; second byModel = %+v", second.byModel)
+	}
+	if first.cost.total == second.cost.total {
+		t.Fatalf("cost cache reused stale model pricing: first=%f second=%f", first.cost.total, second.cost.total)
+	}
+}
+
+func TestReconstructInitialToolCountsKeepsEmptyBranchFallbackArmed(t *testing.T) {
+	reconstructInitialToolCounts(nil)
+	if !needsBranchCountReconstruction() {
+		t.Fatal("empty initial branch must keep before_agent_start reconstruction fallback armed")
+	}
+
+	reconstructInitialToolCounts([]sdk.BranchEntry{{ID: "u1", Role: "user"}})
+	if needsBranchCountReconstruction() {
+		t.Fatal("populated initial branch should satisfy reconstruction")
+	}
+}
+
+func TestReconstructToolCountsFromBranch(t *testing.T) {
+	branch := []sdk.BranchEntry{
+		{Role: "assistant", ToolCalls: []sdk.ToolCallInfo{{Name: "bash"}, {Name: "read"}}},
+		{Role: "user", ToolCalls: []sdk.ToolCallInfo{{Name: "ignored"}}},
+		{Role: "assistant", ToolCalls: []sdk.ToolCallInfo{{Name: "bash"}, {}}},
+	}
+
+	reconstructToolCountsFromBranch(branch)
+
+	mu.Lock()
+	defer mu.Unlock()
+	if totalToolCalls != 4 {
+		t.Fatalf("totalToolCalls = %d, want 4", totalToolCalls)
+	}
+	if toolCallCounts["bash"] != 2 || toolCallCounts["read"] != 1 || toolCallCounts["unknown"] != 1 {
+		t.Fatalf("toolCallCounts = %#v", toolCallCounts)
+	}
+}
+
+func TestEstimateToolDefTokens(t *testing.T) {
+	tools := []sdk.ToolInfo{
+		{Name: "bash", Description: "Execute a command"},
+		{Name: "read", Description: "Read a file"},
+	}
+	tokens, chars, _, total := estimateToolDefTokens(tools)
+	if total != 2 {
+		t.Errorf("totalCount = %d, want 2", total)
+	}
+	wantChars := len("bash") + len("Execute a command") + len("read") + len("Read a file")
+	if chars != wantChars {
+		t.Errorf("chars = %d, want %d", chars, wantChars)
+	}
+	if tokens == 0 {
+		t.Error("tokens should be > 0")
+	}
+}