coding-artifact-evaluation-shell 0.1.0-eidolon.a1b37cbcd7c07 → 0.1.0-eidolon.a3d53b93d02d8

package/capsule.ts

@@ -22,14 +22,12 @@ import {counterReview} from './support/counter-review';
 import {assertCodingScope,type CodingScope} from './coding-scope';
 import {projectTargetGroups} from './target-groups';
 import {retainEvaluation,assertRetentionDestination} from './support/retention';
-import {attachCodeKnowledgeEvidence,type CodeEvidenceSelection} from './support/code-knowledge';
-export {attachCodeKnowledgeEvidence,type CodeEvidenceSelection} from './support/code-knowledge';
 export {createCodexRuntime,type AgentRuntime} from './support/agent-runtime';
 export {retainEvaluation,verifyRetainedEvaluation} from './support/retention';
 export type {CodingScope} from './coding-scope';
 
 /** Runtime dependencies are trusted caller inputs, never discovered in delivered assets. */
-export interface EvaluationOptions {space?:number;judge?:JudgePort;budget?:Budget;noModel?:boolean;experimentalCounterReview?:boolean;agentRuntime?:AgentRuntime;judgeRuntime?:AgentRuntime;codingScope?:CodingScope;retainTo?:string;codeEvidence?:CodeEvidenceSelection}
+export interface EvaluationOptions {space?:number;judge?:JudgePort;budget?:Budget;noModel?:boolean;experimentalCounterReview?:boolean;agentRuntime?:AgentRuntime;judgeRuntime?:AgentRuntime;codingScope?:CodingScope;retainTo?:string}
 interface ObservationRuntime {
   run:AgentRun;agent:AgentRuntime;state:ProcessState;budget:Budget;invocations:Invocation[];
   review?:JudgePort['review'];onProposal?:(proposal:Proposal)=>void;
@@ -104,7 +102,6 @@ export async function runEvaluation(manifest:Manifest,output:string,options:Eval
   if(options.retainTo)assertRetentionDestination(options.retainTo,[output,fs.realpathSync(manifest.artifact)]);
   if(options.experimentalCounterReview&&(options.noModel||options.judge))throw Error('Experimental counter-review requires the native model runtime');
   const run=prepareExecution(bindRuntime(createEvaluationRun(output,manifest.artifact),manifest.runtime),manifest.execution),agent=options.agentRuntime??resolveAgentRuntime('codex'),judgeAgent=options.judgeRuntime??agent,state={interrupted:false};
-  if(options.codeEvidence)manifest=attachCodeKnowledgeEvidence({...manifest,artifact:run.workspace},options.codeEvidence);
   const ceiling=options.noModel?0:(manifest.agentObservation?6:2);
   const budget=options.budget??newBudget(Date.now(),ceiling,1200000);
   const sources=readSources(manifest.targets),artifactDigest=fingerprint(run.workspace),runId=randomUUID();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "coding-artifact-evaluation-shell",
-  "version": "0.1.0-eidolon.a1b37cbcd7c07",
+  "version": "0.1.0-eidolon.a3d53b93d02d8",
   "private": false,
   "type": "module",
   "description": "Code-development artifact evaluator and code-first Halfcode SkillApp",

package/resources/evaluate.md

@@ -25,10 +25,4 @@ For grouped results, the caller explicitly supplies `{id,label,rootIds}` entries
 
 For current coding runs, explicit nonempty `codingScope.groups` also asks the same judge to compare every applicable caller-owned leaf, including after an overall FAIL. It does not add a role, scoring protocol, evidence or new requirements. An absent claim is unreviewed, distinct from an explicit indeterminate claim. Historical and ungrouped minimal FAIL decisions retain their original meaning; completeness and semantic correctness still require actual evidence.
 
-Record the caller's intended scope and method/depth in its trusted codingScope description and capability declarations; report the work actually completed. The current description is retained caller metadata, not an automatic model-input field, execution profile or target filter: authoritative acceptance boundaries belong in the complete original targets, and actual grants/conditions in trusted capability inputs. See [Review](review.md) for the shared evidence-depth guidance used by both collection and judgment. The existing claim method/conditions/reasoning/limitations support scope, actual investigation, coverage and gaps; there is no new wire field or depth score. A code-review-only selection cannot certify an original full-delivery requirement left unobserved. Index/search tools are optional, explicitly supplied locators whose version, resolution and coverage must be checked against frozen source; they do not substitute for execution or unavailable historical facts.
-
-## Optional code navigation evidence
-
-Trusted callers of `runCodingEvaluation` may supply `codeEvidence:{path,sourceRoot}` in options. Both paths are relative to their owned frozen artifact; the evidence JSON must be outside the selected source subtree. This selects the existing file observation channel, not a new judge or execution permission. The runner validates the complete raw-byte snapshot and coverage before a model call, reads the bounded navigation file, and back-reads its bounded UTF-8 candidate source files. With the option omitted, existing behavior is unchanged.
-
-The evidence can be produced by depa-knowledge-base.cs `tools/Depa.CodeEvidence`. It includes unsupported/excluded/oversize files, actual parser/resolvers, unresolved calls and omitted graph counts. Inspect these conditions when reasoning; source locations are candidates, missing hits do not prove absence, and paths establish no runtime or historical event. The adapter authenticates source binding and structure, not semantic correctness of a parser or all requirements. Keep the original complete targets and ordinary source/process/browser observations.
+Before collection, describe the caller's intended scope and method/depth in its trusted codingScope description and capability declarations; report the work actually completed. See [Review](review.md) for the shared evidence-depth guidance used by both collection and judgment. The existing claim method/conditions/reasoning/limitations support scope, actual investigation, coverage and gaps; there is no new wire field or depth score. A code-review-only selection cannot certify an original full-delivery requirement left unobserved. Index/search tools are optional, explicitly supplied locators whose version, resolution and coverage must be checked against frozen source; they do not substitute for execution or unavailable historical facts.

package/support/code-knowledge.ts

@@ -1,73 +0,0 @@
-import * as fs from 'node:fs';
-import * as path from 'node:path';
-import {digest} from './isolation';
-import type {Manifest} from '../contract';
-
-/** Caller-selected, optional navigation. No tool process, graph DB or new judge authority. */
-export interface CodeEvidenceSelection {path:string;sourceRoot:string}
-interface ByteIdentity {path:string;sha256:string;bytes:number}
-export function sourceByteInventory(root:string):ByteIdentity[] {
-  const result:ByteIdentity[]=[];let bytes=0;
-  function visit(dir:string):void {
-    if(fs.lstatSync(dir).isSymbolicLink())throw Error('Code evidence cannot follow links');
-    for(const name of fs.readdirSync(dir)){
-      const file=path.join(dir,name),stat=fs.lstatSync(file);
-      if(stat.isSymbolicLink())throw Error('Code evidence cannot follow links');
-      if(stat.isDirectory()){visit(file);continue;}
-      if(!stat.isFile())throw Error('Code evidence requires regular files');
-      const relative=path.relative(root,file).split(path.sep).join('/');
-      if(/[\0\r\n\\]/.test(relative))throw Error('Unsupported snapshot path');
-      bytes+=stat.size;if(result.length>=10000||bytes>128*1024*1024)throw Error('Code snapshot exceeds inventory bound');
-      result.push({path:relative,sha256:digest(fs.readFileSync(file)),bytes:stat.size});
-    }
-  }
-  visit(root);return result.sort((a,b)=>a.path<b.path?-1:a.path>b.path?1:0);
-}
-export function sourceByteDigest(files:ByteIdentity[]):string {
-  return digest(files.map(f=>`${f.path}\0${f.sha256}\0${f.bytes}\n`).join(''));
-}
-function owned(root:string,relative:string):string {
-  if(!relative||path.isAbsolute(relative)||relative.split(/[\\/]/).some(p=>p==='..'||p==='.'||p===''))throw Error('Code evidence path escapes artifact');
-  let current=root;
-  for(const part of relative.split('/')){
-    current=path.join(current,part);
-    if(fs.lstatSync(current).isSymbolicLink())throw Error('Code evidence cannot follow links');
-  }
-  return current;
-}
-
-/** Called after native artifact freezing; stale envelopes fail before model invocation.
- * The ordinary file observer reads this envelope plus bounded candidate source files.
- * Unsupported/unreadable sources remain in inventory; no observation is invented. */
-export function attachCodeKnowledgeEvidence(manifest:Manifest,selection:CodeEvidenceSelection):Manifest {
-  const root=fs.realpathSync(manifest.artifact),source=owned(root,selection.sourceRoot),file=owned(root,selection.path);
-  if(file===source||file.startsWith(source+path.sep))throw Error('Evidence must be outside its source snapshot');
-  if(fs.statSync(file).size>1048576)throw Error('Code evidence exceeds file observation bound');
-  const evidence=JSON.parse(fs.readFileSync(file,'utf8'));
-  if(evidence.format!=='depa-code-navigation-v1')throw Error('Unknown code evidence format');
-  const inventory=sourceByteInventory(source),identity=sourceByteDigest(inventory);
-  if(evidence.snapshotDigest!==identity||JSON.stringify(evidence.inventory)!==JSON.stringify(inventory))throw Error('Stale or different code evidence snapshot');
-  if(!Array.isArray(evidence.coverage)||evidence.coverage.length!==inventory.length)throw Error('Incomplete code evidence coverage');
-  const seen=new Set<string>(),fileIds=new Map<string,string>();
-  for(const row of evidence.coverage){
-    const original=inventory.find(f=>f.path===row.path);
-    if(!original||seen.has(row.path)||original.sha256!==row.sha256||original.bytes!==row.bytes||!['document','text-or-regex-fallback','semantic-candidate','over-size-or-excluded','skipped','unsupported-or-excluded'].includes(row.status)||!Array.isArray(row.resolvers))throw Error('Invalid code evidence coverage');
-    seen.add(row.path);if(row.fileId)fileIds.set(row.fileId,row.path);
-  }
-  if(!Array.isArray(evidence.symbols)||!Array.isArray(evidence.edges)||evidence.symbols.length>200||evidence.edges.length>400||!evidence.parser||!evidence.settings||!Array.isArray(evidence.limitations)||!evidence.limitations.length)throw Error('Incomplete bounded code evidence');
-  for(const kind of ['symbols','edges'] as const){
-    const omitted=kind==='symbols'?'omittedSymbols':'omittedEdges';
-    if(!Number.isSafeInteger(evidence.totals?.[kind])||evidence.totals[kind]<evidence[kind].length||evidence.totals[omitted]!==evidence.totals[kind]-evidence[kind].length)throw Error('Invalid code evidence truncation');
-  }
-  for(const row of [...evidence.symbols,...evidence.edges])if(row.fileId&&!fileIds.has(row.fileId))throw Error('Code location outside frozen inventory');
-  const probes=[...(manifest.files??[])];
-  const add=(relative:string,conditions:string)=>{if(!probes.some(p=>p.path===relative))probes.push({path:relative,conditions});};
-  add(selection.path,'Optional static code navigation from byte-bound frozen source. Inspect coverage, parser/resolvers, unresolved calls and truncation. Missing hits prove no absence; paths prove no execution or past event. Back-read source before using a candidate; original requirements and judge authority are unchanged.');
-  for(const relative of new Set<string>(evidence.symbols.map((s:{fileId:string})=>fileIds.get(s.fileId)).filter(Boolean))){
-    const content=fs.readFileSync(path.join(source,relative));
-    try{if(content.length>1048576)continue;new TextDecoder('utf-8',{fatal:true}).decode(content);}
-    catch{continue;}
-    add(selection.sourceRoot+'/'+relative,'Actual frozen source read to check optional navigation candidates. Static inspection establishes neither runtime execution nor completeness of unsupported files.');
-  }
-  return {...manifest,files:probes};
-}