Files
yoloyolo/.planning/phases/03-pipeline-integration-and-mvp/03-01-PLAN.md
T
2026-03-26 13:02:16 +01:00

405 lines
16 KiB
Markdown

---
phase: 03-pipeline-integration-and-mvp
plan: 01
type: execute
wave: 1
depends_on: []
files_modified:
- classify/types.go
- classify/classifier.go
- classify/classifier_test.go
- synth/config.go
- synth/bank_test.go
- synth/config_test.go
autonomous: true
requirements: [CLAS-02]
must_haves:
truths:
- "AllClasses() returns exactly 14 classes with no ClassUnknown"
- "Unrecognized TCP/UDP packets are deterministically assigned to ClassUnknown1-4 via hash-bucketing"
- "Packets with no transport layer (ARP etc.) are assigned to an unknown bucket, not silently dropped"
- "Each unknown bucket has a distinct tone in the 850-1100 Hz dissonant range"
- "GainPerLayer is 1/14 so 14 layers mix without clipping"
artifacts:
- path: "classify/types.go"
provides: "ClassUnknown1-4 constants, 14-element AllClasses()"
contains: "ClassUnknown1"
- path: "classify/classifier.go"
provides: "hashBucket() function routing unrecognized traffic"
contains: "func hashBucket"
- path: "synth/config.go"
provides: "FreqConfig entries for all 14 classes, NumLayers=14"
contains: "NumLayers = 14"
- path: "synth/config_test.go"
provides: "Test ensuring every AllClasses() entry has a FreqConfig"
contains: "TestClassFreqConfigsComplete"
key_links:
- from: "classify/classifier.go"
to: "classify/types.go"
via: "hashBucket returns ClassUnknown1-4 constants"
pattern: "ClassUnknown[1-4]"
- from: "synth/config.go"
to: "classify/types.go"
via: "ClassFreqConfigs map keys include all 14 classes"
pattern: "classify\\.ClassUnknown[1-4]"
---
<objective>
Extend the traffic classification type system from 11 classes to 14 by replacing ClassUnknown with 4 hash-bucketed unknown classes, and add corresponding synthesis tone configurations.
Purpose: Implements CLAS-02 (auto-clustering) so unrecognized traffic produces distinct drone tones instead of collapsing into a single undifferentiated layer.
Output: Updated classify and synth packages with 14-class type system, hash-bucketing classifier, dissonant tone configs, and comprehensive tests.
</objective>
<execution_context>
@$HOME/.claude/get-shit-done/workflows/execute-plan.md
@$HOME/.claude/get-shit-done/templates/summary.md
</execution_context>
<context>
@.planning/PROJECT.md
@.planning/ROADMAP.md
@.planning/STATE.md
@.planning/phases/03-pipeline-integration-and-mvp/03-CONTEXT.md
<interfaces>
<!-- Key types and contracts the executor needs. -->
From classify/types.go (CURRENT — will be modified):
```go
type TrafficClass string
const (
ClassICMP TrafficClass = "ICMP"
ClassDNS TrafficClass = "DNS"
ClassHTTPS TrafficClass = "HTTPS"
ClassHTTP TrafficClass = "HTTP"
ClassSSH TrafficClass = "SSH"
ClassSMTP TrafficClass = "SMTP"
ClassNTP TrafficClass = "NTP"
ClassDHCP TrafficClass = "DHCP"
ClassOtherTCP TrafficClass = "other-TCP"
ClassOtherUDP TrafficClass = "other-UDP"
ClassUnknown TrafficClass = "unknown" // TO BE REMOVED
)
func AllClasses() []TrafficClass // currently returns 11, will return 14
```
From classify/classifier.go (CURRENT — ClassUnknown return sites at lines 23, 37, 52, 67, 71):
```go
func (c *Classifier) Classify(pkt gopacket.Packet) ClassifiedPacket
// Falls through to ClassUnknown when no rule matches
```
From synth/config.go (CURRENT — will be modified):
```go
const NumLayers = 11
const GainPerLayer = 1.0 / float64(NumLayers) // ~0.0909
var ClassFreqConfigs = map[classify.TrafficClass]FreqConfig{...} // 11 entries
```
From encode/mp3.go (NOT modified — consumes AllClasses() via synth.NewBank):
```go
func RunSynthesis(snapshots []classify.WindowSnapshot, outputPath string) error
```
</interfaces>
</context>
<tasks>
<task type="auto" tdd="true">
<name>Task 1: Extend TrafficClass type system and write Wave 0 tests</name>
<files>classify/types.go, classify/classifier_test.go, synth/config_test.go, synth/bank_test.go</files>
<read_first>classify/types.go, classify/classifier_test.go, synth/bank_test.go, synth/config.go</read_first>
<behavior>
- TestAllClassesCount: `len(AllClasses()) == 14`
- TestAllClassesNoPlainUnknown: no element in AllClasses() equals `"unknown"`
- TestAllClassesHasUnknownBuckets: AllClasses() contains `"unknown-1"`, `"unknown-2"`, `"unknown-3"`, `"unknown-4"`
- TestHashBucketDistribution: calling hashBucket with ports 0-999 and protocol "tcp" produces all 4 ClassUnknown1-4 values
- TestClassifyUnknown (UPDATE): buildUnknownPacket returns a class with `strings.HasPrefix(string(got.Class), "unknown-")`, not `ClassUnknown`
- TestNewBankHas11Layers (UPDATE): assert `len(b.layers) == 14` (rename to TestNewBankHas14Layers)
- TestClassFreqConfigsComplete (NEW in synth/config_test.go): every class from `AllClasses()` has a key in `ClassFreqConfigs`
</behavior>
<action>
**Step 1: Update classify/types.go (per D-04)**
Remove the `ClassUnknown TrafficClass = "unknown"` constant. Add four new constants:
```go
ClassUnknown1 TrafficClass = "unknown-1"
ClassUnknown2 TrafficClass = "unknown-2"
ClassUnknown3 TrafficClass = "unknown-3"
ClassUnknown4 TrafficClass = "unknown-4"
```
Update `AllClasses()` to return 14 elements:
```go
func AllClasses() []TrafficClass {
return []TrafficClass{
ClassICMP, ClassDNS, ClassHTTPS, ClassHTTP, ClassSSH,
ClassSMTP, ClassNTP, ClassDHCP, ClassOtherTCP, ClassOtherUDP,
ClassUnknown1, ClassUnknown2, ClassUnknown3, ClassUnknown4,
}
}
```
**Step 2: Add tests to classify/classifier_test.go**
Add `TestAllClassesCount` subtest:
```go
func TestAllClassesCount(t *testing.T) {
classes := classify.AllClasses()
if len(classes) != 14 {
t.Errorf("AllClasses() returned %d classes, want 14", len(classes))
}
// No plain "unknown" should exist
for _, c := range classes {
if c == "unknown" {
t.Error("AllClasses() still contains plain \"unknown\" — should be removed per D-04")
}
}
// All 4 buckets must be present
buckets := map[classify.TrafficClass]bool{
classify.ClassUnknown1: false, classify.ClassUnknown2: false,
classify.ClassUnknown3: false, classify.ClassUnknown4: false,
}
for _, c := range classes {
if _, ok := buckets[c]; ok {
buckets[c] = true
}
}
for bucket, found := range buckets {
if !found {
t.Errorf("AllClasses() missing bucket %q", bucket)
}
}
}
```
Add `TestHashBucketDistribution` (will fail until Task 2 implements hashBucket — that's the TDD red phase, both tasks are in the same plan so this is fine):
```go
func TestHashBucketDistribution(t *testing.T) {
// hashBucket is unexported, so test through Classify with unmatched TCP ports
c := classify.NewClassifier(classify.DefaultRules)
seen := make(map[classify.TrafficClass]bool)
// Try a range of unmatched TCP ports to hit all 4 buckets
for port := uint16(10000); port < 11000; port++ {
pkt := buildTCPPacket(t, port)
got := c.Classify(pkt)
if strings.HasPrefix(string(got.Class), "unknown-") {
seen[got.Class] = true
}
}
for _, bucket := range []classify.TrafficClass{
classify.ClassUnknown1, classify.ClassUnknown2,
classify.ClassUnknown3, classify.ClassUnknown4,
} {
if !seen[bucket] {
t.Errorf("hashBucket never produced %q across ports 10000-10999", bucket)
}
}
}
```
Add `"strings"` to the import block in classifier_test.go.
Update `TestClassifyUnknown` subtest (line 220-226): replace the assertion with:
```go
t.Run("TestClassifyUnknown", func(t *testing.T) {
pkt := buildUnknownPacket(t)
got := c.Classify(pkt)
if !strings.HasPrefix(string(got.Class), "unknown-") {
t.Errorf("Unknown packet: got class %q, want unknown-N bucket", got.Class)
}
})
```
**Step 3: Update synth/bank_test.go**
In `TestNewBankHas11Layers` (line 10-21):
- Rename function to `TestNewBankHas14Layers`
- Change assertion from `!= 11` to `!= 14`
- Change error message from `want 11` to `want 14`
In `TestMixerNoClip` (line 76-102):
- Update comment from `All 11 classes` to `All 14 classes`
- Update `TotalPackets` from `11000` to `14000`
**Step 4: Create synth/config_test.go**
New file:
```go
package synth
import (
"testing"
"github.com/netsynth/netsynth/classify"
)
func TestClassFreqConfigsComplete(t *testing.T) {
for _, class := range classify.AllClasses() {
if _, ok := ClassFreqConfigs[class]; !ok {
t.Errorf("ClassFreqConfigs missing entry for class %q", class)
}
}
}
func TestNumLayersMatchesAllClasses(t *testing.T) {
if NumLayers != len(classify.AllClasses()) {
t.Errorf("NumLayers=%d but AllClasses() has %d entries", NumLayers, len(classify.AllClasses()))
}
}
```
</action>
<verify>
<automated>cd /home/dev/workspace/yoloyolo && go vet ./classify/... ./synth/... 2>&1</automated>
</verify>
<acceptance_criteria>
- classify/types.go contains `ClassUnknown1 TrafficClass = "unknown-1"` through `ClassUnknown4 TrafficClass = "unknown-4"`
- classify/types.go does NOT contain `ClassUnknown TrafficClass = "unknown"` (plain unknown removed)
- classify/types.go AllClasses() returns slice with 14 elements
- classify/classifier_test.go contains `func TestAllClassesCount`
- classify/classifier_test.go contains `func TestHashBucketDistribution`
- classify/classifier_test.go TestClassifyUnknown uses `strings.HasPrefix(string(got.Class), "unknown-")`
- synth/bank_test.go contains `func TestNewBankHas14Layers` (not 11)
- synth/bank_test.go TestNewBankHas14Layers asserts `!= 14`
- synth/config_test.go contains `func TestClassFreqConfigsComplete`
- synth/config_test.go contains `func TestNumLayersMatchesAllClasses`
- `go vet ./classify/... ./synth/...` exits 0 (compiles cleanly)
</acceptance_criteria>
<done>Type system extended to 14 classes, all Wave 0 tests written, code compiles. Some tests will fail (RED phase) until Task 2 implements hashBucket and updates config.</done>
</task>
<task type="auto">
<name>Task 2: Implement hash-bucketing and update synth config for 14 layers</name>
<files>classify/classifier.go, synth/config.go</files>
<read_first>classify/classifier.go, synth/config.go, classify/types.go, classify/classifier_test.go, synth/config_test.go, synth/bank_test.go</read_first>
<action>
**Step 1: Add hashBucket function to classify/classifier.go (per D-01/D-03)**
Add this unexported function at the bottom of classifier.go:
```go
// hashBucket maps an unrecognized packet to one of 4 unknown traffic classes.
// Deterministic: same (dstPort, protocol) always maps to the same bucket.
// ARP and other non-transport packets have dstPort=0, protocol="" -> bucket 0 (ClassUnknown1).
func hashBucket(dstPort uint16, protocol string) TrafficClass {
var protoNum uint16
switch protocol {
case "tcp":
protoNum = 6
case "udp":
protoNum = 17
case "icmp":
protoNum = 1
}
h := uint32(dstPort)*31 + uint32(protoNum)*7
switch h % 4 {
case 0:
return ClassUnknown1
case 1:
return ClassUnknown2
case 2:
return ClassUnknown3
default:
return ClassUnknown4
}
}
```
**Step 2: Replace all ClassUnknown return sites in Classify() method**
There are 4 sites where ClassUnknown is used in classifier.go:
1. Line 23 — initial default: change `ClassUnknown` to `ClassUnknown1` (temporary; this gets overwritten by the final fallthrough, but the initial value doesn't matter since every code path sets Class before returning. However, to avoid a stale reference, use `""` or just leave the field unset since the default zero value is fine).
Actually, the simplest approach: change the initial default at line 23 to empty string `""`, then ensure every return path sets the class. The 4 return paths are:
a. **ICMP no-rule-match (line 37):** `return result` after ICMP rule loop fails. Replace with:
```go
result.Class = hashBucket(result.DstPort, result.Protocol)
return result
```
b. **TCP no-rule-match (line 52):** `return result` after TCP rule loop. Replace with:
```go
result.Class = hashBucket(result.DstPort, result.Protocol)
return result
```
c. **UDP no-rule-match (line 67):** `return result` after UDP rule loop. Replace with:
```go
result.Class = hashBucket(result.DstPort, result.Protocol)
return result
```
d. **No transport layer (line 71):** `return result` at the very end. Replace with:
```go
result.Class = hashBucket(result.DstPort, result.Protocol)
return result
```
Remove the initial `Class: ClassUnknown` at line 23. Instead set `Class: ""` or omit the field (zero value for string is ""). The `Classify` doc comment referencing "Per D-03: returns ClassUnknown" should be updated to "Per D-03: returns ClassUnknown1-4 via hash-bucketing".
**Step 3: Update synth/config.go (per D-02/D-05/D-06)**
Replace `NumLayers = 11` with `NumLayers = 14`.
This automatically changes `GainPerLayer = 1.0 / float64(NumLayers)` to ~0.0714.
Remove the `classify.ClassUnknown` entry from `ClassFreqConfigs`. Add 4 new entries:
```go
// D-05: Unknown buckets in 850-1100 Hz dissonant range, detuned intervals
// D-06: Same dissonant harmonic character {1,1.0},{2,0.8},{3,0.4} for all 4
classify.ClassUnknown1: {862.0, []HarmonicDef{{1, 1.0}, {2, 0.8}, {3, 0.4}}, 0.6},
classify.ClassUnknown2: {920.0, []HarmonicDef{{1, 1.0}, {2, 0.8}, {3, 0.4}}, -0.6},
classify.ClassUnknown3: {981.0, []HarmonicDef{{1, 1.0}, {2, 0.8}, {3, 0.4}}, 0.9},
classify.ClassUnknown4: {1047.0, []HarmonicDef{{1, 1.0}, {2, 0.8}, {3, 0.4}}, -0.9},
```
Remove the old `classify.ClassUnknown` entry and its comment about 437 Hz / D-04 beating.
</action>
<verify>
<automated>cd /home/dev/workspace/yoloyolo && go test ./classify/... ./synth/... -count=1 -v 2>&1 | tail -40</automated>
</verify>
<acceptance_criteria>
- classify/classifier.go contains `func hashBucket(dstPort uint16, protocol string) TrafficClass`
- classify/classifier.go does NOT contain the string `ClassUnknown` without a digit suffix (no bare ClassUnknown references)
- classify/classifier.go hashBucket uses `uint32(dstPort)*31 + uint32(protoNum)*7` and `h % 4`
- synth/config.go contains `NumLayers = 14`
- synth/config.go contains `classify.ClassUnknown1:` through `classify.ClassUnknown4:`
- synth/config.go ClassUnknown1 BaseHz is 862.0, ClassUnknown2 is 920.0, ClassUnknown3 is 981.0, ClassUnknown4 is 1047.0
- synth/config.go does NOT contain `classify.ClassUnknown:` (old entry removed)
- `go test ./classify/... -count=1` passes (all classifier tests green including TestHashBucketDistribution, TestAllClassesCount, TestClassifyUnknown)
- `go test ./synth/... -count=1` passes (TestNewBankHas14Layers, TestClassFreqConfigsComplete, TestNumLayersMatchesAllClasses, TestMixerNoClip all green)
</acceptance_criteria>
<done>Hash-bucketing classifier routes unrecognized traffic to 4 distinct unknown classes. Synth config has tone entries for all 14 classes. All classify and synth tests pass.</done>
</task>
</tasks>
<verification>
```bash
# All classify and synth tests pass
go test ./classify/... ./synth/... -count=1
# Full suite still passes (no regressions in encode, aggregate, capture)
go test ./... -count=1
# No bare ClassUnknown references remain (should only find ClassUnknown1-4)
grep -rn 'ClassUnknown[^1234]' classify/ synth/ encode/ cmd/ aggregate/ | grep -v '_test.go.*unknown-' | grep -v '// '
```
</verification>
<success_criteria>
- AllClasses() returns 14 classes; ClassUnknown is fully removed
- hashBucket deterministically maps (dstPort, protocol) to ClassUnknown1-4
- All 4 unknown buckets have FreqConfig entries in 850-1100 Hz range
- GainPerLayer = 1/14; TestMixerNoClip passes with 14 layers
- All tests in classify/ and synth/ pass
- No bare ClassUnknown references remain in production code
</success_criteria>
<output>
After completion, create `.planning/phases/03-pipeline-integration-and-mvp/03-01-SUMMARY.md`
</output>